Migration Pressure, Renewable

Resource Scarcity, and Internal Armed

Conflict

Oda Fjeldvær Eggen

Master‟s Thesis in Political Science

UNIVERSITY OF OSLO

May 2010

II

III

© Oda Fjeldvær Eggen

2010

”Migration Pressure, Renewable Resource Scarcity, and Internal Armed Conflict.”

Oda Fjeldvær Eggen

http://www.duo.uio.no/

Print: Oslokopisten

IV

V

Acknowledgements

Although writing this Master‟s Thesis at times has been an incredibly lonely process in

a way, there are those that have made it less so. First of all my mom and dad deserve

gratitude for being unconditionally supportive as always, and reminding me that I‟m

the best no matter what (apologies to my brother). I am also very grateful to my

wonderful friends and my big brother, who have been indispensible in keeping my

motivation up; and staying in school. My boyfriend, Kai Arne, definitely deserves

thanks for being surprisingly understanding, for forgiving my at times horrible moods,

and for making my day.

The professors at the University have all been incredibly helpful, particularly Knut-

Andreas Christophersen with his door always open, and Håvard Hegre; however

intimidating he might seem at first.

Last but not least I would like to thank my supervisor, Henrik Urdal, for being patient,

motivating, and for always making time for me. It may be obvious, but; I couldn‟t

have made it without him!

Remaining errors are my responsibility alone.

Number of words (all included): 25 225

VI

VII

Contents

1.0. Introduction ....................................................................................................................... 1

2.0. Theory ................................................................................................................................ 7

2.1. Scarcity and conflict ........................................................................................................ 8

2.2. Migration and Resource Scarcity .................................................................................. 19

2.3. The Impact of Migration on Different Types of Violence ............................................ 27

2.4. Migration and International Aid .................................................................................... 29

3.0. Research Design and Data .............................................................................................. 33

3.1. Previous Empirical Studies ........................................................................................... 34

3.2. Method: Logistic Regression ........................................................................................ 36

3.3. Time-Series Regression Analysis .................................................................................. 37

3.4. Operationalization of variables ..................................................................................... 38

3.3.1. Dependent Variable ................................................................................................ 38

3.3.1.1. Violent Conflict ............................................................................................... 38

3.4.2. Independent Variables ............................................................................................ 40

3.4.2.1. Migration Pressure .......................................................................................... 40

3.4.2.2. Renewable Resource Scarcity ......................................................................... 41

3.4.2.3. International Aid ............................................................................................. 44

3.4.3. Control Variables ................................................................................................... 44

3.4.4. Validity and Reliability .......................................................................................... 46

4.0. Time-Series Analysis ....................................................................................................... 49

4.1. Conventional Conflict ................................................................................................... 51

4.2. Non-State Conflict ......................................................................................................... 55

4.3. Analysis Summary ........................................................................................................ 58

5.0. Conclusion ........................................................................................................................ 60

6.0. Bibliography .................................................................................................................... 65

7.0. Appendix .......................................................................................................................... 79

7.1. Descriptive Statistics over the Dependent Variables .................................................... 79

7.2. Variable Correlation Matrix .......................................................................................... 79

VIII

7.3. Logistic Regression Time-Series Results in Coefficients ............................................. 81

7.4. “Do-File” (STATA 9 ) .................................................................................................. 81

List of Tables

Table 1: Conventional Conflict Logistic Regression Results, 1964-2007 (Odds Ratio) ... 51

Table 2: Conventional Conflict Logistic Regression Results, 1950-2001 (Odds Ratio) ... 52

Table 3: Non-State Conflict Logistic Regression Results, 2002-2007 (Odds Ratio) ......... 56

1

1.0. Introduction

Environmental issues are increasingly on the international agenda. The rising

awareness of problems created by global environmental degradation has been spurred

by, amongst other things, the introduction of the concept of sustainable development

by the International Union for the Conservation of Nature and Natural Resources

(IUCN) in 1980, and gaining currency with the Brundtland report of 1987. Numerous

scholars have, without reaching consensus on the issue, suggested or denied

connections, between the fight for access to natural resources and the rise of tension

and conflict (Hauge & Ellingsen 2001: 37). Relating environment and conflict, seems

to be of growing interest. Many theoretical assumptions have been made about the

relationship between environmental degradation and conflict; however, few empirical

studies on the topic have given significant support to these suppositions. Quantitative

large-N studies in particular, struggle to find strong connections between violent

internal conflict and renewable resource scarcity (Urdal 2005; Theisen 2006). An

important aspect of the environmental context is population, and migration is a crucial

part of this.

Migration has been connected both to conflict and the environment mainly by the

much-debated concept of “environmental refugees”. Myers‟ predictions of a displaced

population of at least 200 million people due to environmental strains have had great

influence on the debate concerning such refugees. He has identified an emergent

problem that, if correct, necessarily is of great significance (1997: 168). It is

speculated that this phenomenon represents one of the potentially greatest human

crises of our time, and that it is an “outward manifestation of profound change – a

manifestation often marked by extreme deprivation, fear and despair” (Myers 1997:

181). The most widely repeated predictions of 200 million refugees are, however,

contested, and the repetition in itself does not make the figure any more accurate

(Brown 2007: 2). These refugees are claimed to be displaced because of environmental

causes, whether abrupt or more subtle, but all due to environmental change.

2

Categorizing these people as such is in itself problematic, and may be misleading.

Estimates of their number vary, as do their definition. Environmental change has

always been a natural part of human adaptation to such change, and still is, a necessity.

As a way of adjusting to these changes, people have historically migrated over smaller

or larger areas to improve their predicaments. Although the consequences of

environmental change might have accelerated, it is difficult to compare expected

consequences to the extraordinary. Some downplay the urgency of the migration

problematic by claiming that the migration effects of environmental distress may be

absorbed by the urbanization trend. Such evolvement may, however, in turn pose even

larger challenges to potential local conflict levels.

Population displacement due to environmental degradation and change is an issue, but

so is the impact of population growth on the environment, due to its potentially violent

consequences. There are, however, conflicting views on the magnitude of population

growth, and the projections vary accordingly. Recent history has arguably rejected the

most extreme predictions, and there are even those who claim that the population

growth is so much in decline that one claim to have “turned a corner on population

growth” (Goldstone 2001: 96). The reasons for this may be an increase in education

for women, the spread of economic development, and /or national and international

support for policies of population planning (and accompanying movement along the

demographic transition), which in turn has contributed to dropping fertility and

population growth rates in different parts of the world. There is, however, no denying

a considerable overall population increase. Considering moderate estimates of a rise to

9 billion people by 2050, and assuming international migration to remain at no more

than a 3% of the world‟s population, migration would increase to approximately 275

million in 2050. This growth is expected to be unevenly distributed, with less

developed regions being particularly affected. Estimates suppose a global doubling of

the age group 15-24, which traditionally are those most prone to migrate (Black et al.

2008). The consequences of population growth and the additional population pressure

of migration is an obvious challenge. It will be crucial to establish an understanding of

possible consequences of such likely future scenarios, particularly in a resource scarce

3

context. Weiner and Teitelbaum argue that “Population – its growth or decline, its

movement, its density, its characteristics, its distribution – has always been linked to

questions of security.” This entails that the movement of peoples has transformed

societies, and both made and unmade states. In addition, population growth has been

regarded a source of national power, or alternately been a contributor to disorder and

violence. Contemplation following this same logic has grasped the attention of ancient

and classical political philosophers and commentators (2001: ix). Environmental issues

have also been incorporated into security thinking, and have at times been labelled

“the new security issue”. This perspective is clarified by the reasoning of Myers:

If a nation’s environmental foundations are depleted, its economy will steadily

decline, its social fabric deteriorate, and its political structure become

destabilized. The outcome is all too likely to be conflict, whether conflict in the

form of disorder and insurrection within the nation, or tensions and hostilities

with other nations (Myers 1986: 251).

Threats to security are thus being more broadly defined today than in the past.

Environmental security is an unconventional source of threat, though the threat of

conflict remains a conventional concern. Homer-Dixon provoked great concern and

controversy when claiming that we are on the verge of an era where environmental

change will more frequently trigger conflict, if not be the primary source of insecurity

(Goldstone 2001: 84-88). Despite the consequential security thinking, Homer-Dixon

himself argues that the environment-security theme incorporates an almost

unmanageable range of sub-issues, especially if the security term is broadly defined.

He recommends a narrowing down of the research question to primarily investigating

conflict rather than security, however still claiming the topic being too vast (1991: 76-

77). Even though armed conflict in the world today is at a historic low, and that there

has been a decline in conflict throughout the post-war era, recent years show an

escalation in armed conflict world wide. This recent increase is suggested due to

intrastate rather, than interstate conflict (Eriksson & Wallensteen 2004: 625). Several

4

scholars agree that our understanding of the potential security risk that is posed by

environmental degradation is increasingly complex and are marked by a pattern of

conflicts within countries, rather than interstate conflicts (Ohlsson1999; Deudney

1990; Homer-Dixon 1999). Ohlsson argues that in the wake of the genocide in

Rwanda in 1994, the Director of USAID, Brian Atwood, noting that issues of natural

resources are frequently critical to achieving political and economic stability, while

also identifying disintegrated societies and failed states as the greatest menace to

global stability, took the first step of identifying the new pattern of conflicts in the

world (Ohlsson 1999: 26).

Identifying environmental degradation as a security issue is in its own right no given.

Not all scholars agree that this status is well earned. Barnett argues that environmental

change and its impact on social and ecological systems, together with a post cold war

fluid international security environment, encourages environmental degradations to be

viewed as a security concern (2001: 2). Although emphasising interstate conflict,

Deudney admits that given the trend of deteriorating environment as a consequence of

the accelerating race toward higher standards of living, environmental issues are likely

to “become an increasingly important dimension of political life at all levels” (1990:

461). However, he severely doubts the link between environmental degradation and

national security. Deudney claims that this upgrade to a national security concern is

owed primarily to the traditional focus on national security, and it being a main feature

when it comes to both intrastate and interstate conflict (1990: 461). Suhrke has an

almost normative approach to the question of environmental security. When referring

to the history of migration, she argues that migrants and refugees may represent

victims, assets, or threats. By selecting the most negative dimension “the security

paradigm is likely to reinforce popular stereotypes of migrants as undesirable and

dangerous, even if the analyst does not draw this conclusion” (1997: 256, see also

Weiner 1993; 1995). This does not render the security paradigm obsolete in this

context, but requires that it is treated with caution, particularly when it comes to the

suitability to empirical material (Suhrke 1997: 256).

5

This Master‟s thesis starts from the neo-Malthusian assumption that demographic

change (mainly population growth) and resource scarcity, are important sources of

armed conflict. Demographic factors have arguably been linked to conflict from time

immemorial. Demographic aggression is thus by no means a new concept, although

demographic variables and forms of demographic aggression vary over time and place

(Weiner and Teitelbaum 2001: 46). In this thesis, the main demographic variable of

interest is migration pressure, operationalized as refugees, and the context is that of

resource scarcity. Measuring migration pressure through refugees captures sudden

increases in population relative to the access of renewable resources in a better way

than by looking at general population density or growth. The supposition is that sharp

increases in competition over resources caused by sudden increase in migration will

place heavy burdens on the physical environment.

The forms of aggression studied are both low and high intensity conflicts, and include

non-state and state actors. A broadened definition of conflict avoids a limitation to

conventional conflict, and provides a broader test of the neo-Malthusian argument.

There are theoretical reasons for assuming increased occurrence of minor rather than

major conflicts as a result of environmental factors (Homer-Dixon 1995: 12)1. It can

also be argued that the most likely type of conflict to emerge is one internal to

countries. This entails that the supposed intensity of a conflict can be lower both due to

an expectation that the actors involved are less organized and equipped than those in

conventional conflicts, and that the context of the conflict, suffering from scarcity of

important resources, possibly represents strain on conflict activity in its own right2.

This is accounted for in this thesis by defining conflict at a low intensity level (25

1 Homer-Dixon and Blitt investigates both state- and non-state conflict, but does not

noteworthy distinguish between the two. In the publication “Ecoviolence” the conventional

conflicts in Gaza, Chappas and Rwanda are investigated, as are the non-state conflicts of

Pakistan and South Africa (1998).

2 The indirect mortality rate can nonetheless be quite high under these circumstances. Conflict

in an already precarious environmental context can have fatal consequences for civilians and

combatants alike (e.g. Darfur).

6

battle-related deaths per year) than what is commonly used in civil-war studies (1000

battle deaths), and by including non-conventional conflict, where the state is not an

active party to the conflict. The study will take the form of a large-N quantitative

cross-national time-series study, covering the period 1950-2007. The unit of analysis is

the country-year. The dependent variable is armed conflict onset. Migration pressure,

renewable resource scarcity and international presence, are all central independent

variables. The main focus is thus; the effects of migration on conflict in a resource

scarce context. The causal relationship between migration and scarcity is assumed to

have a mutually reinforcing positive effect on the outbreak of conflict. This study is

restricted to domestic conflict, and does not consider interstate conflict. International

aid is expected to have a mediating, or at least negative effect on this relationship. The

focal point of the thesis is reflected in the following general hypothesis, later to be

specified in two sub-hypotheses:

Migration pressure has a greater impact on the probability of conflict the greater

the level of renewable resource scarcity.

The interaction of migration and resource scarcity in potentially causing conflict, have

so far not been thoroughly investigated, leaving room for the contribution of this

Master‟s Thesis. The focus on migration pressure where resources are scarce, and its

implications on the risk of conflict, is an interesting topic. Migration, as well as

resource scarcity, is by many understood as increasing, and it is important to explore

whether the conjunction of these two phenomena is likely to spur conflict. To

investigate these possible connections, the analysis will as mentioned use a

quantitative design. Although few large-N studies find strong relationships between

renewable resource scarcity and internal conflict, this thesis may be justified both by

the alternative causal connections it investigates, and by the use of relatively new and

diverse data, particularly considering the inclusion of low-intensity and non-state

conflict.

7

2.0. Theory

„Environmental refugees‟ seem to be naturally associated with the subject of

environmental degradation and migration. This concept has become progressively

noticeable, and the term continues to gain way both in the contemporary climate-

debate and among scholars. The estimated number of environmental refugees is

disputed, as are the definitions of who exactly qualifies to constitute this population.

Estimating the magnitude of environmental migration is almost impossible, or difficult

at best (Suhrke 1993: 33). The definitional boundaries of refugee-generating

environmental changes have varied, ranging from “sudden natural disasters to gradual

natural disasters to human-induced disasters.” (Lee 2001: 55) Although there seems to

be little doubt that environmental disruption contributes to large population

displacement, how to determine its scale with certainty remains a challenge (Lee 2001:

55-58). There is an admission of the fact there is no “mono-causal relationship

between climate change, disasters and displacement” (NRC 2009: 5). Although, it is

recognized that there does exists a link between the phenomena. The distinction

between involuntary and voluntary migration is controversial, yet essential (Suhrke

1991: 9). This distinction is used to contrast environmental refugees from migrants.

Where environmental refugees respond to push factors, migrants act based on a

combination of pull and push factors. Refugees, thereby, flee relatively unprepared

compared to migrants (Suhrke 1991: 9-10). While voluntary migration is an obvious

adaptation strategy, climate change and disasters may also cause forced displacement

that is essential for survival (NRC 2009: 5). United Nations Development Programme

(UNEP) researcher El-Hinnawi first defined environmental refugees as:

[…] those people who have been forced to leave their traditional habitat,

temporarily or permanently, because of a marked environmental disruption

(natural and/or triggered by people) that jeopardized their existence and/or

seriously affected the quality of their life [sic]. By ‘environmental disruption’ in

this definition is meant any physical, chemical, and/or biological changes in the

8

ecosystem (or resource base) that render it, temporarily or permanently,

unsuitable to support human life. (El-Hinnawi 1985: 4).

Environmental changes may lead to different kinds of migration, depending on the

characteristics of these changes, and resulting in acute versus slow onset of

environmentally induced migration (Zolberg and Benda 2001). The urgency of

migration due to extreme weather may lead to more temporary displacement than

displacements provoked by slower processes, such as rising sea levels. State capacity

may influence the mere scale of the migration, as well as possibilities for repatriation.

Bear in mind that when discussing refugees in this study, these are categorized as such

in the more traditional sense of the word.

2.1. Scarcity and conflict

Many theoretical arguments derive from the original work of Thomas Malthus, and

connect population pressure to conflict. This postulates the simple, but profound,

notion that while food production grows linearly, population increases tends to be

exponential (Tir & Diehl 2001: 61). “All of us who ponder the questions of the human

environment are the intellectual descendants of Thomas Robert Malthus”, no matter if

one should agree on his teachings or not (Kates 1996: 45). The theoretical framework

of this thesis will explore the neo-Malthusian presumption of the unsustainability of

food-production and population growth. Contemporary understanding of population

and resource issues can be said to derive from three principal dimensions of concern:

Malthusian, Marxist and Ecological. Malthus assumed that the number of people

would grow at a geometric rate3, thereby overwhelming the resources that are

3 “Geometric growth refers to the situation where successive changes in a population differ

by a constant ratio (OECD)” [Emphasis added].

9

expanding only at an arithmetic rate4. Continuations of this demographic pessimism

are the Marxists. Although refuting the Malthusian premise, this view defines the

population problem in terms of distribution. It is argued that existing resources will be

sufficient, if only they are properly utilized and distributed. This will be achieved

through a transformation from capitalism to socialism. The Ecologists brought forth

the implications of ecological factors on the political and economic failure or success

of a nation. Meadows and Meadows’ book The Limits to Growth (1972) was a

benchmark in this regard. This perspective challenges the inevitability of continued

industrialization, from where most ecological pressure stems (Lee 2001: 8-9). Political

ecologists consider resource-related conflicts to be primarily motivated by structural

inequalities, rather than “natural” scarcities and population growth (Kahl 2006: 23).

Environmental depletion is assumed to further hurry these incompatible developments.

Many environmental scarcity models widely rely on the assumptions of relative

deprivation theory. This implies that renewable resource scarcity will encourage socio-

economic grievances that eventually result in conflict (Theisen 2008: 814-815). This

is further specified by Ohlsson, who defines the general understanding of the

Malthusian dilemma as:

[...] great and global processes of change that follow from continuing large

population increases, rising developmental expectations, the unavoidable

environmental impacts that follow as these expectations gradually are realized,

and the consequences, in turn, for people as their societies are put to such

severe strains resulting from these large processes of change that their very

ability to fulfil the undeniable right to better lives for their populations is

threatened, ultimately entailing risk of violent conflict (Ohlsson 1999: 3).

The works of Homer-Dixon and the Toronto Group are placed within the relative

deprivation tradition in conflict theory. The core concept of Homer-Dixon‟s work is

4 Distinct from the concept above. Increase by a constant number (OECD).

10

environmental scarcity, distinguished by three dimensions. The supply-induced

scarcity dimension implies that the availability of a renewable resource is reduced

faster than its regeneration. Demand-induced scarcity results from growth in

population or increase in per capita consumption. Structural scarcity occurs where

resources are inequitably distributed and rest in possession of a privileged minority,

while the remaining population suffers from resource shortage (Homer-Dixon 1994).

Environmental scarcity is a product of these parts working in conjunction. The

Toronto Group emphasises that any treatment the fundamental issue of resource

renewable resource scarcity should “encompass the exhaustive set of scarcity‟s

resources: decreases in supply, increases in demand and changes in distribution

(Schwartz, Deligiannis and Homer-Dixon 2001. 276).” These different categories of

scarcity interact and are mutually reinforcing, resulting in two social processes

designated „resource capture‟ and „ecological marginalization‟. The latter occurs when

population groups faced with resource scarcity migrate into areas with fragile eco-

systems, which in turn generates greater scarcities in the given area as well as

deprivation conflicts between locals and newcomers (Homer-Dixon 1999). Concerning

„resource capture‟, there is arguably a social inequality effect caused by society‟s

powerful, resulting in ecological marginalization of the less advantaged (Ohlsson

1999: 38).

None of these theoretical assumptions, however, go unchallenged. The Malthusian

analysis has been somewhat discredited mainly as a consequence of remarkable

augmentation in food production and our capacity to both exploit and transform our

environment (Weiner and Teitelbaum 2001: ix). Julian Simon represents one of the

leading advocates of the school of thought that does not consider population increase

to be as dangerous. By this he challenges the basic assumption that population growth

inevitably leads to resource scarcity. The argument is that most resources are not

genuinely fixed in the economic sense, and are therefore not likely to diminish due to

rising population. Resources are valuable only for the „services‟ they provide. Because

the resources we exploit are perpetually changing, resources are not finite. The

ultimate scarce resource is claimed to be human ingenuity (1981; 1996). There are

11

nevertheless difficulties concerning this “optimist” view. First of all, it largely discards

the lag-time between population pressure and necessary technological adjustments,

and the fact that these will vary across countries and regions. Secondly, technological

innovation is both unequally distributed and accessible in time and space. Finally, the

assumption that technological advancement will be the salvation to population growth

is argued to be “neither sufficiently justified theoretically nor entirely confirmed

empirically” (Tir & Diehl 2001: 65-66).

Political economy as a field, including the views of Malthus, has been accused of

tending to “reduce everything to social construction, blatantly disregarding all that is

not human” (Greenberg & Park 1994: 1). Political ecologists generally claim to

consider both cultural and political activity within an analysis of ecosystems,

considered to be significantly, yet not entirely social constructs. For political

ecologists „the environment‟ ranges from the predominantly cultural, through the

intensely political to the considerably natural (ibid: 8). Peluso and Watts reject

automatic and oversimplified linkages between increased environmental scarcity,

decreased economic activity, migration, and the resulting violence and conflict that

Homer-Dixon and his colleagues (the Toronto group) are accused of. They consider

violence to be the product of local histories and social relations, while still very much

being part of larger processes of power relations and material transformation. The

importance of patterns of accumulation is emphasised. These patterns are formed by a

structured political economy, which differentiates positions of, and access to,

resources (Peluso & Watts 2001: 5). Peluso and Watts refuse scarcity to have

monopoly over violence, which they consider a claim of Homer-Dixon‟s. Both

resource scarcity and abundance alike can provoke violence. They call for more

thorough investigation of the ways in which environmental violence disguises and

reflects other forms of social struggle (ibid: 5-8). Political ecology generally argues

that environmental issues become „socialized‟ when collective resources are controlled

by local groups at the expense of others. This again forces management interventions

by private firms, state agents or development authorities. By the same logic, conflicts

12

within and between communities are „ecologized‟ by conservation or resource

development policies (Robbins 2004: 14). Elements of environmental issues are

increasingly present, but hidden by the perpetrators and observers of violence equally.

The legacy of Malthus and the like is considered more of a curse than anything. Peluso

and Watts understand the predictions and causalities behind environmental scarcity

and conflict as provoked by a “deep fear of the poor and their claims to resources,

despite radical changes in the world since Malthus‟ time (Peluso & Watts 2001: 5-8).

Studies conducted by Homer-Dixon and the Toronto Group, have been under heavy

scrutiny (Gleditsch & Urdal 2002; Gleditsch 1998). It is claimed that when these

investigate the relationships between environment and conflict, there is little or no

variation in neither the dependent nor the independent variable. This is because the

literature is mainly based on case studies, where both stress on the environment and

armed conflict are or have been present. This makes generalizations difficult (Hauge

and Ellingsen 2001: 37-40). Some argue that most empirical findings from

environment-conflict case analyses acknowledge an interaction and mutual stimulation

of conflict and environment, and thereby conclude that little, if any, conflict can be

strictly defined as environmental conflict (Lee 2001). Another problem is that the

concept of environmental scarcity in itself is based on other factors than environmental

degradation. The idea of structural scarcity, for instance, is concerned with the unequal

distribution of resources and is in large part a product of politics. The politics of

distribution is accused of disappearing in the overarching environmental scarcity

concept (Hauge and Ellingsen 2001: 37-40).

Appreciating the complexity of the environment-conflict nexus, Homer-Dixon restricts

his research ambitions. Rather than evaluating the whole spectre of independent

variables, he weighs the seriousness of the processes or pathways whereby the

environment may affect the degree of dependent variables. Despite these precautions,

Lee states that while the environment is merely one of many causally significant

13

factors in the complex conditions of conflict, it is still reasonable to claim that “with

intensifying environmental decline, violent conflict involving environmental

components has been noticeably increasing” (2001: 5-6). The Toronto group in

general, is accused of failing to contribute to the understanding of the causal pathway

to domestic armed conflict by ignoring more direct linkages between these conflicts,

and political and economic factors. Gleditsch‟s (1998) critique of Homer-Dixon and

the Toronto group is mainly rooted in methodological concerns, such as inappropriate

selection of cases, the neglect of the possibility for reverse causation, devising

untestable models, lacking tools to weigh causal variables and overemphasising the

complexity of ecological-political systems. This critique to Homer-Dixon represent a

“methodological straightjacket that would, if widely adopted, severely constrain

research in the field” (Schwartz, Deligiannis and Homer-Dixon 2001: 273-274). It is

further argued that these deep methodological issues “can only be understood in the

context of Gleditsch‟s unduly narrow perspective on what constitutes systematic

research (Schwartz, Deligiannis & Homer-Dixon 2001: 280)”. Homer-Dixon and

colleagues defend their methodology by emphasizing the difference between causal

effect and causal mechanism. Neither the single-case nor experimental and quasi-

experimental method is sufficient in revealing any causation with absolute certainty.

The single-case method is even so a necessary instrument to demonstrate causation

(Schwartz, Deligiannis & Homer-Dixon 2001: 282).

More fundamentally, the relative importance of renewable resource scarcity and

depletion remains debated (Theisen 2006: 5). The focus on the degradation of

renewable resources is, on a theoretical level, reinforced by the concept of the “tragedy

of the commons”. This illustrates a mechanism whereby natural resources, not yet

subjected to social management or regulation, apparently inevitably are depleted due

to an increase in demand, particularly from population growth. This tragedy occurs

mainly in periods of rapid change, when social systems‟ capacity to adapt has proven

insufficient, or when regimes of resource management have been challenged by

competing modes of production (Ohlsson 1999: 18). Gurr even argues that scarcity

14

seldom is an objective fact, but rather is a social construct shaped by people‟s

perception, given that scarcity rarely exists as an absolute fact. “Scarcity as a social or

political problem is largely defined by people‟s perception of the lack of a resource in

terms of their image of the good life” (Welch & Miewald 1983: 10). However, Gurr

argues that ecological scarcity could breed gradual impoverishment, creating material

inequalities both within and among societies. These inequalities are, in turn, claimed to

intensify class cleavages, ethnic hostilities and conflicts (1985). Defining scarcity like

this diverts somewhat from the more commonly used economic conception of resource

scarcity (Gurr 1985: 55). Ohlsson also stresses the relative nature of 'scarcity'.

Although a seemingly straightforward concept, indicating “a situation where there is

insufficient amount of a particular resource or asset to satisfy normal requirements”.

Determining what is 'normal' and how that norm has evolved, however, is highly

problematic. Considering the concept of scarcity as a social construct, and to

acknowledge it as such, is to “recognize the impact of cultural change on material

needs” (1999: 3-4).

The very urgency of resource depletion problematic is scrutinized by the notion of

human abilities of adaptation. Ingenuity is important in order to be able to handle

resource scarcity and environmental degradation successfully. The ongoing debates

present diverting opinions on what may inspire the necessary creativity. Boserup

(1965) argues that in agriculturally dependent societies, the mere pressure on resources

will necessarily lead to both innovation and economic diversification. de Soysa further

reasons that the abundance of resources (be that non-renewable or renewable)

obstructs incentives for innovation. This is because states become less dependent upon

skilled labour by collecting the majority of its revenues from resource rents, and

thereby under-develop human capital, considered to be the most important determinant

for economic growth (de Soysa 2002, 2005). As a result of ever increasing resource

consumption, population growth and resource access, resource substitution and

conservation tasks will become ever more urgent, significantly increasing the need for

ingenuity (Homer-Dixon 1999: 26). Societies need to adapt by ingenuity, and both

15

understand and act on links between environmental scarcity and violence, and its

negative social effects (Homer-Dixon 1999: 107). With necessity being the mother of

all invention, there will be an assumed increase in ingenuity that will help alleviate

scarcity‟s severity and social impacts (Homer-Dixon 1999: 108).

Homer-Dixon admits that the answers concerning what affects a society‟s ability to

supply much needed ingenuity, is highly complex. Ingenuity is relative to many

factors, such as social, political, economic and cultural characteristics of the different

countries, and these respond to scarcity in different ways. Homer-Dixon claims that

with a rise in both population and per capita resource consumption, in addition to

persistent inequalities in resource access; renewable resource scarcity will affect

especially poor countries with “unprecedented severity, speed and scale” (Homer-

Dixon 1999: 26). This need is however challenged by influences and circumstances

such as limited access to capital, brain drain, incompetent bureaucracies, generally

weak states and corrupt judicial systems. In addition, the supply of ingenuity can be

limited by stresses generated by the very resource insufficiency it envisioned to solve.

There is said to be an “ingenuity gap” where the “requirement for ingenuity to deal

with environmental scarcity rises while their supply of ingenuity stagnates or drops”

(Homer-Dixon 1999: 26-27).

The argument continues that in poorly governed countries, and especially those with a

history of conflict, there is also a lack of incentives to invest both time and labour in

the conservation of resources. Combined with low government interest and investment

in rural areas, bad governance is held to be the causal mechanism behind both resource

degradation and civil war (de Soysa 2002, 2005). Other scholars also emphasize

government resources and the quality of institutions as key determinants of whether or

not increasing scarcity will lead to conflict (Kahl 2006). Raleigh and Urdal argue that

compared to all economic, political and social factors, environmental and demographic

stress is not likely to be an equally important risk factor. The first mentioned factors

are considered to not only determine the state capacity to adapt, but also largely mould

16

the general opportunities for rebel groups to succeed (2007: 676).

Gleditsch argues that politics have a mediating effect on the environment and conflict

relationship5. Everything else being equal, democracies have a more “enlightened

environmental behaviour”. Democracies are claimed to be more responsive to those

affected by environmental degradation, and are also more prone to actively cooperate

internationally in order to alleviate environmental problems (2001: 57). The

importance of state capacity is stressed to be a determinant of the impact of migration.

Kahl recognises two causal pathways from demographic and environmental stress,

understood as resource scarcity, and violent conflict6. These causal arguments are

„state-centric‟, and form the hypothesis of state exploitation and state failure. The

latter meaning a weakening of the state. This in turn heightens the potential gain from

a rebellion compared to the likely costs a state is able to inflict on the rebels, and

thereby an increase risk of conflict is to be expected (ibid 2006: 47). Demographic and

environmental stress may negatively interfere with a state‟s functional capacity by

costly demands placed to remedy strains on the agricultural sector, and needs for social

improvements (Kahl 2006: 40-43).

State capacity may be considered a reflection of its vulnerability. Vulnerability is by

the Intergovernmental Panel on Climate Change (IPCC) defined as “the degree, to

which a system is susceptible to, and unable to cope with, adverse effects of climate

change, including climate viability and extremes” (2001: 995). This concept thereby

captures both the risk and degree of exposure, as well as the ability to handle the

5 Gleditsch also claims that politics influence the very manner in which conflicts are acted

out. This reasoning stems from the concept of democratic peace (See: Gleditsch & Hegre

1997; Raknerud & Hegre 1997). This concept is however considered to lie outside the scope

of this Thesis.

6 Armed conflict is throughout this thesis defined with a lower threshold of 25 annual battle-

deaths, as used in the Uppsala dataset (Gleditsch et al., 2002; datasets available at

http://www.prio.no/cwp/datasets).

17

challenges imposed by the environment. Social cohesion may be put to the test by

diverting understandings of the best response to challenges of resource scarcity, put

forward by competing parties. An undermining of overall economic activity may

generally contribute negatively to both a states functional capacity and its abilities of

social cohesion (Kahl 2006: 40-43). Even when demographic and environmental

factors are not the drivers behind state failure, such factors contribute to a relative

weakening of the state and these are considered more perceptible to conflict over

resource scarcity. This increased risk of conflict is both due to the states inability to

mitigate effects of resource scarcity, and because they are overall more likely to be

militarily challenged by opposition groups (Raleigh & Urdal 2007: 679). While large

migrations stem from environmental problems, a future migration crisis is equally a

crisis of social, political and economic sorts, that mirrors institutional systems‟

inabilities to effectively reflect unique and deep-seated changes, such as this (Myers

1997: 181).

Homer-Dixon does in fact acknowledge the importance of context. Environmental

scarcity is never a sole or sufficient cause of violence, poverty or large migrations.

Environmental scarcity is assumed to always interact with other economic, political

and social factors to produce the social effects described above (Homer-Dixon 1999:

16). This implies that for conflict to break out as a result of environmental factors, it

takes more than solely environmental degradation, or resource scarcity. The adaptive

capacity of the state in particular and society in general, is crucial (Ohlsson 1999: 48).

Given an environmental issue, however, “there must be both an opportunity and a

clearly perceived advantage of taking to arms” (ibid: 48). As claimed by Homer-

Dixon, countries experiencing rapid growth in population are dependent on adaptation

to be able to avoid conflict in comparison to countries with a high but stable

population-to-resource ratio (Homer-Dixon 1999). Moreover, Homer-Dixon explicitly

emphasises the heightened possibility for conflict outbreak within, rather than between

states. “These environmental scarcities do not cause wars among countries, but they

18

can generate severe social stresses within countries, helping to stimulate sub national

insurgencies, ethnic clashes, and urban unrests” (Homer-Dixon 1999: 12).

Hauge and Ellingsen have found that economic and political factors are the strongest

predictors of conflict, but that environmental and demographic factors nonetheless

have impact (1998). Kahl addresses this by arguing that by way of economic,

ecological and social effects, population and environmental pressure reverberate into

politics that by state failure and state exploitation, turn into pathways to civil strife

(2006: 29-30). The state may instigate conflict in their advantage by focusing on

migrants. Rejecting the migrant population may be used in order to gain popularity and

support from the domestic population. Suhrke argues that this role may be reversed by

the state aligning itself with the displaced. Such scenarios are considered more likely

in situations where the support of a displaced group is determined on grounds of ethnic

politics or economics (Suhrke 1993: 28). This underlines the alleged critical

relationship between “the migrant, the refugee and the state”. The potential for acute

conflict thereby lies in the migrant population‟s ability to gain support and make

demands (Suhrke 1997: 270).

In a quite opposite scenario, and in the absence of state intervention, refugees

generated and driven by environmental pressures tend to be victims, rather than

threats. These migrants are weakened, disempowered and dispersed, and pose no

immediate threat. They are claimed to make few effective demands on the receiving

area, and their most urgent needs are commonly met by international relief. Social

exploitation, rather than blatant social conflict, takes place (Suhrke 1997: 270).

Raleigh and Urdal emphasise difficulties in testing the opportunity aspect of a state

exploitation. As derived from Kahl‟s (2006) concepts of such a potential role of a

state, one can imagine a state instrumentally using resource scarcity as a means to

bolster support. Encouraging (e.g. inter-ethnic) conflict over resources, may contribute

to divert attention from state incapability of meeting domestic demands. Though

theoretically appealing, Raleigh and Urdal argue that the hypothesis is too vaguely

19

defined to be able to isolate effects and test by statistical analysis (2007: 679-680).

Homer-Dixon is also aware that environmental degradation in itself is an insufficient

cause of violence, recognizing the importance of economic, political and social factors.

He is, however, concerned that scholars may “swing to the opposite extreme” of what

he himself been accused for, namely underestimating important contextual factors.

First of all, environmental scarcity can itself affect social factors like institutions and

policies in harmful ways, and not only the other way around. Secondly, mere physical

characteristics of the society‟s surrounding environment can be a factor. Lastly, once

environmental scarcity becomes irreversible, it is as if by definition, an external

influence on society (Homer-Dixon 1999: 17).

2.2. Migration and Resource Scarcity

Migration has naturally always existed, but scholars seem to agree that there, for

different and controversial reasons, has been an increase at least in forced migration,

across and within international border (Wood 1994; Castles & Miller 2005). Millions

of migrants and refugees, and their hopes for freedom from violence and repression are

coincidently matched by the fears of states and their citizens, that a massive influx of

newcomers will impose strains on the economy, upset a possible precarious ethnic

balance, weaken national identity or threaten political upheaval (Weiner and

Teitelbaum 2001: 107-108). Whether or not these governmental concerns are

justifiable is difficult to assess, although perhaps not from the lack of trying.

Comprehensive and generalisible theories are considered to remain elusive. Linkages

between demography and security, and their predictive implications, have been linked

to theory, but intervening variables between causes and consequences, highly

contextual outcomes and data limitations, are accused of confounding empirical

analysis (Weiner and Russell 2001: 16).

20

Cause and effect relationships between environmental degradation and migration are

difficult to quantify, and are considered tied to political, cultural and economic factors

(Zolberg and Benda 2001: 44). Migration may be both a cause and effect of worsening

environmental conditions. There may be overlapping environmental, economic and

political push factors, as well as pull factors in the receiving area determining whether,

when and where to move (Buhaug et al: 2008: 21). Although the causes of migration

are not pivotal to this particular study, the points made are nevertheless useful.

Reasons to flee can be many. Traditionally, as acknowledged by the United Nations,

refugees are those who migrate in fear of being persecuted because of “race, religion,

nationality, membership of a particular social group, or political opinion (UNHCR

1951: Article 1)”. This definition assumes across-border migration, as does the

operationalization of migration pressure in this Thesis‟ analysis. The widely accepted

definition of a refugee, is thereby someone fleeing from war or conflict, having

crossed an internationally acknowledged border and usually granted political asylum.

Distinguishing a refugee from a voluntary migrant is imperative.

It is the reluctance to uproot oneself, and the absence of positive original

motivations to settle elsewhere, which characterises all refugee decisions and

distinguishes the refugee from the voluntary migrant (Kunz1973: 130 in Hugo

1996: 109).

There are two different forms of migration that are considered somewhat intertwined.

Voluntary migration is motivated by several influences derived from economic,

political and ideological reasoning, but also considering environmental factors. Forced

migration may stem from direct environmental factors that create unbearable living

conditions, but this degradation may in its own turn be a product of underlying

economic and political factors. Following this same logic, Unruh, Krol and Kliot more

precisely define that migration triggered from conflict caused by resource depletion,

“does not occur because of the direct consequence of environmental change but rather

as a result of a complex series of interlinked (“snowballing”) factors in which single

clear-cut cause-to-effect relations may not be identifiable” (2004: VIII). In such a

perspective, the distinction between economic and environmental refugees is not

21

straightforward. Further blurring these particular motivations is the notion of eco-

migrants; “eco” stemming from both the term ecology and that of economy. Eco-

migrant is an even more indistinct concept than the “environmental refugee”. Eco-

migrants include those voluntarily moving to new areas in order to exploit natural

resources. However, these same people are often forced to leave, as resources on

which they depend are severely degraded or destroyed (Zolberg and Benda 2001: 47).

Many scholars have attempted to nuance migration. Buhaug, Gleditsch and Theisen

differentiate between rapid and gradual, and permanent or temporary migration. These

differentiations are influenced by the speed of the perceived environmental pull and

push factors. A further distinction is made by identifying those who flee from

immediate dangers, and separate from those who travel over longer distances with

hopes of a better future in a different area (2008: 27). When discussing complex issues

of what can lead to migration of populations, Unruh, Krol and Kliot claim an

important distinction between voluntary and forced migration (2004: VIII). Olson

added the dimension of physical danger, which is environmentally induced, to the

established premise of persecution. He defines refugees, and thereby forced migration,

as follows:

Refugees differ from other, spontaneous or sponsored migrants, largely in the

circumstances of their movement out of one area to another, and the effects

these have on them in the settlement and adjustment phases of their relocation.

Refugees are forced to leave their homes because of a change in their

environment which makes it impossible to continue life as they have known it.

They are coerced by an external force to leave their homes and go elsewhere

(1979: 130 in Hugo 1996: 107).

22

Salehyan also differentiates types of migration, distinguishing between environmental

migrants and „classic‟ refugees. It is claimed that environmental refugees do not have

political agendas, unlike refugees that flee from conflict zones and that are inclined to

make political demands and have an interest in the conflict outcome of their native

countries. If environmental degradation leads to conflict and thereby forces migration,

then these refugees are assumed to have a greater propensity to provoke conflict in the

receiving area (2005: 13). There have been several studies that consider the nature of

migrants, whether being environmentally induced or driven by conflict, to be a

determinant on the risk of conflict in a receiving country (Forsberg 2009a; Salehyan

2005; Buhaug & Gleditsch 2008). However, it is far from obvious that the reasons for

migration will have significant security implications for the host area. Lack of

conceptual clarity and data limitations has restricted the opportunity to empirically

study the possible impacts of environmental migration across cases (Buhaug et al.

2008: 28). What constitutes a refugee might not be sufficiently defined through formal

specifications, but presumes a subjective understanding of the situation.

Motives for flight are normally associated with conflict. Although present in earlier

literature, the influence of environmental factors on refugee flows, has received

increasing attention. One may distinguish between the environment as primary

objective for fleeing, and the environment being an accessory interacting with other

motivating factors. Unruh, Krol and Kliot speak of forced migration as being “one of

the direct or indirect effects of global environmental change (2004: VII).” They further

argue that such migration, leading to what they call environmental refugees most

likely will have significant social, economic and political consequences. They predict

a political and economic tension raised explicitly by an increase in the number of

refugees that in turn may lead to conflict situations. This conflict potential spurs from

most contemporary governments ability to deal with such a situation (ibid). Zolberg

and Benda claim that the primary push factor is political, and not environmental (2001:

46). Weiner argues that there are further trends that demonstrate an increase in internal

conflicts, much due to ethnic conflict. Wars between states remain a diminishing, but

23

nevertheless significant source of refugee flows. It is also found that the number of

refugees produced increases far more rapidly than the number of countries producing

refugees, implying an increase in refugees per conflict. This increase is claimed to be a

consequence of the natural population increase in countries of origin, the availability

of arms on both sides of a conflict lowering the bar of taking to arms, and the

increased use of antipersonnel mines causing menace toward the population that

persist even past the original conflict‟s end (1996: 6, 25-26).

Lee argues that, as hypothesised from a neo-Malthusian perspective, high population

density within nations may per se cause social disintegration and, at the extreme,

increase the risk of violent conflict over limited resources. Additional population

pressure in already overpopulated areas, such as Third World countries are often

aggravated by large-scale flows of migrants and refugees (Lee 2001: 11-13).

Understandings of the term „overpopulation‟ is however not a given. This stems from

the acknowledgement of the fact that there is no intuitively ideal ratio between

population and resources. Limiting overpopulation to the Third World is also

misleading, as developing countries may well be far less densely populated, and still

have the capacity to absorb these. Goldstone insists on the security implications of

population change, even with a possible decrease in population growth (Goldstone

2001: 96-97). The importance of agriculture and access to resources is also

emphasised:

Many countries may well experience collisions between their agrarian

populations and access to land, between the expansion of their labour force,

educated aspiring elites, urban population, and youth cohorts and the

absorption rate of their economies, and […] between migrants and resident

populations that inflame ethnic and regional tension (Goldstone 2001: 99).

24

Concerns of both population growth and resource constraints are rejected by Boserup

(1981) who holds that these are vital to promote technological progress.

„Cornucopians‟ such as Boserup and Simon, reject the Malthusian assumption that

population promotes resource scarcity and promotes civil strife. Concerning

agriculture for instance, Boserup expects a parallel progression in agricultural

technology as pressure on the land increases (Boserup 1981). Simon (1996) argues that

resource depletion is avoided through the technological process. It seems an increase

in population is understood as an increase in people to innovate.

Migration is at large considered to be an intermediate stage linking environmental

degradation to conflict (Homer-Dixon 1991; 1994). One may speculate that refugee

populations increase conflict risk given a scenario of an overwhelming of local

services, and provoking violence and resentment in receiving areas. Suhrke, however,

claims that such expectations belong in the realm of local fears, rather than in social

reality (Suhrke 1993: 34). She claims it is a common misperception that conflict will

ensue when people are displaced, and argues that this can only be the case in zero-sum

interaction, however actual or perceived (1997: 257-258). Weiner and Teitelbaum,

however, see a conflict potential proper to refugee populations. There is a real concern

that massive flows of refugees offer a setting where aggressive states and non-state

actors may strategically „place‟ their operative‟s sensitive locations abroad (2001:

108). Also Forsberg underlines such conflict risk assigned to refugees in particular,

investigating possible contagion effects between refugees and neighbouring countries

(2009a). Looking at the spread of conflict trough migration-flows is encouraged by

evidence suggesting that ethnic conflicts, and thereby also ethnic groups, are not

customarily confined within the borders of one particular state, implying transnational

linkages between ethnic groups involved in conflict and group members living in a

neighbouring country (Forsberg 2009a: 25-26).

25

Even though developed societies are less susceptible to environmental stress triggered

by increased population pressure, most refugees migrate between developing

countries. In such areas, relatively small numbers of refugees may present significant

challenges, generating enormous pressures. Given the magnitude of global

environmental damage, the contribution to environmental degradation caused by

refugees may be minimal. Even so, unexpected increases in population may be a

challenge to the ecological balance in the affected area, and this in turn may generate

economic and social strains (Lee 2001: 110-112). Taking this reasoning even further,

one may contemplate that in extreme these „strains‟ may result in violent conflict.

Where the number of refugees is high, and their stay prolonged, refugees increase the

rate of resource consumption and depletion, which accelerates environmental

degradation (ibid). This leads to intensified competition between the native and the

newly arrived populations over scarce land and resources, and may again lead to

further migration. The conclusion drawn from this strain of thought is that

environmental and/or political crises producing cross-border refugees may well

generate “other refugee-producing environmental changes and/or conflicts in a

receiving nation” (ibid).

Dramatic growth in cross-border population movement manifests itself all over the

world. This growth mostly occurs between developing countries, often already strained

on resources. There are conflicting perceptions of exactly what may happen after

refugees arrive in a host country. Lee distinguishes between “combatant” and “non-

combatant refugees”. Other than obvious conflict-generating qualities of combatant

refugees, the last category raises challenges of its own. The non-combatant refugees

may arguably become politicized and build their own community, and thereby

accelerating existing internal instability in the host state, when the possibility of

returning home in the near future is slim (2001: 109). This prediction differs

significantly from that of those who believe that a longer stay in the host country will

stimulate more or less harmonic integration in the receiving state.

26

Reuveny identifies at least four complementary processes relating what he calls

“climate-induced migration” to conflict. These four processes leading from migration

to conflict are competition by the burdening of the economic and resource base in the

receiving area; ethnic tension both between migrants and residents or by ethnic divides

despite nationality; distrust between the host and receiving area; and fault lines

following existing socioeconomic patterns (Reuveny 2007: 659). In identifying these

processes to receiving areas that are particularly prone to conflict, Reuveny stresses

the impact and potential tension caused by migration. It is not unreasonable to suggest

that such strains on receiving areas may expect levels of unrest, if not straight out

conflict. In Reuveny study, which has a self acclaimed Malthusian taste, it is nuanced

that although not to overrule the possibility of conflict, climate-induced migration does

not have to lead to conflict. Migration can even benefit the absorbing area by e.g.

increasing the workforce and tax-base (2007: 660). Lee argues along the same lines,

that the political balance in a state or area may be rocked by the multiple consequences

of changes in demographics resulting from migration (Lee 2001: 14). She argues that

in multiethnic and heterogenic societies burdened by conflict among contending

linguistic, ethnic, or religious groupings, foreign populations may disturb domestic

equilibrium. The argument is that given the crucial part demographic factors play in a

democratic process and the political framework of “one person, one vote”, locals may

believe to lose political domination over their own land (ibid.). The Swiss research

project „Environment and Conflict‟ (ENCOP) also recognizes the conflict potential in

migration and environmental scarcity. This project claims that demographically

induced conflict appears when and where there are clear contradictions between

economic and ecological carrying capacity. Indicators of this become visible through

“shrinking per-capita allotments of arable land”. This is also the underlying

assumption in this Thesis‟ empirical analysis, and the operationalization of the

resource scarcity variable. The first hypothesis explores the relationships of migration

pressure, resource scarcity and violent conflict.

Hypothesis I: Migration pressure is particularly likely to increase the risk of

conflict in the context of increasing renewable resource scarcity.

27

2.3. The Impact of Migration on Different Types of Violence

It is claimed to have been found empirical evidence relating both to the positive causal

relationship between environmental scarcity and conflict, but also to the very nature of

conflict. Homer-Dixon claims that environmental scarcity rarely leads directly to

interstate conflict, as states are unable to quickly and with ease convert renewable

resources into assets that importantly augment their power. In addition, the countries

that depend the most on renewable resources tend to be poor and therefore have fewer

capabilities for aggression (Homer-Dixon 2001; 1994; 1999). Scarcities limit already

fragile societies‟ ability to cope with negative changes in the access of resources, by

the additional stress this imposes, and thereby reducing both agricultural and economic

activity, ultimately weakening the state (Homer-Dixon & Blitt, 1998: 280f). Thereby;

“environmental scarcity is mainly an indirect cause of violence, and this violence is

internal to countries (Homer-Dixon 1999: 18).” In this context, conflict is

characterized as tending to be “persistent, diffuse, and sub-national.” This type of

conflict is anticipated to increase intensively in the coming decades, under the

assumption that scarcities rapidly worsen in several parts of the world (Homer-Dixon

1994: 39). Smaller, rather than larger conflicts are considered more interesting in this

setting (Homer-Dixon 1994; 1995). This claim has found empirical bearing besides the

authors of this claim. Hauge and Ellingsen, after operationalizing Homer-Dixon‟s

Environmental Scarcity, demonstrates that environmental degradation seems to have a

stronger impact on the incidence of smaller rather than larger armed conflicts (1998).

The increased possibility of relatively local conflicts at low-intensity levels may in the

long run have severe implications. Salehyan and Gleditsch claim that periods of low-

intensity conflicts, in practice, are “likely to be associated with a higher likelihood of

future large-scale conflict (2006: 350).”

Fighting a government army requires great organizational efforts and considerable

resources. From this perspective, it may be argued that conflicts between non-state

actors (non-state conflicts) are more likely to occur, given that these parties are

affected to a greater extent by environmental scarcities, compared to a given

28

government. The potential underground involvement of a government should however

not be underestimated (Theisen & Brandsegg 2007: 6)7. The traditional focus on

interstate and civil wars when investigating resource scarcity and conflict is criticized.

The argument is that there is an “inverse relationship between the importance of scarce

resources and the scale of conflict”. This implies a call of attention to lower-level

conflicts, not presuming the state as an active part (ibid 2007: 2). Given that non-state

conflict is understood as a breach of the state monopoly of the use of force, the event

of such conflicts must somehow indicate a silent permission from the state, or an

inability to prevent it (ibid: 6). Theisen assumes that in a conflict over resources, one

should expect that non-state conflicts be most likely, relative to those where the state is

an active part. On the premise that scarcity leads to grievances and poverty, deprived

groups will be too weak to engage in conflict with state forces (2006: 35). Suliman

argues that scarcity is most important to more local and smaller conflicts, where the

state is not necessarily an active part. These „local in clinch‟ have a more peripheral

status relative conflicts of the state, as these conflicts are less likely to evolve into full-

scale civil wars (1999b). Given the state‟s neglect of economically less significant

areas, the void of a mitigating third party is filled by local rule. Historically

cooperative bonds between groups are weakened in parallel to the increase of scarcity.

Upsurges of conflict between neighbouring groups can be caused by e.g. the „desert

versus the oasis syndrome‟, where one group inhabits a relatively more fertile eco-

zone than the other (ibid. 1999b: 187). These premises result in the following

hypothesis, implying the nature of conflict.

Hypothesis II: Violent conflict resulting from the interaction between migration

pressure and renewable resource scarcity, is of a low-intensity nature, and does

not presuppose government involvement.

Suhrke emphasises that where displacement becomes long-term, the conflict potential

rises in parallel to the migrant population‟s acquired autonomy or influential

7 Although, this is beyond the scope of this thesis.

29

allegiances. A prevail over the powerlessness inherent with the refugee or migrant

status, will allow an increase in demand to their hosts. Also, gradual migration is

expected to have potentially violent outcomes, and as such, migration may have a

cumulative effect that may place additional stress on already fragile political and social

systems (1997: 263-264). The neo-Malthusian logic of resource depletion is followed

when emphasising that population size and density, as well as individual consumption,

escalate the shortage of resources and environmental degradation. With increasingly

scarce resources, more expensive and environmentally damaging processes are

required in order to provide resources for additional people. Such ecological stress can

launch fierce competition among the affected population, and possibly lead to violent

conflict. Lee argues that there are several stages a society may go through before

resorting to violence. To be able to clarify the evolution of such conflicts enables the

anticipation of “a new dimension of global insecurity”, according to Lee (2001: 21).

2.4. Migration and International Aid

There is an articulated need for the international community to mitigate and adapt to

possible large-scale displacement (Brown 2007: 29). There is reason to believe that the

mere presence of a third party may have a calming effect on a country‟s conflict level.

This will, of course, to large extent also depend on the kind of presence. One could

contemplate that a strong economic presence of third parties would provide an

economic incentive to keep domestic peace, out of consideration for foreign

investments. United Nations agencies and Non-Governmental Organizations (NGOs)

may have more explicit interests in peace, as these third parties are used to obtain

exactly this and mediate between conflicting parties. By explicit mandates aimed at

both preventing and restraining conflict, they already have attentiveness, and an alert

concerning precarious situations. In particular, humanitarian refugee assistance may be

of importance for lowering domestic conflict levels, not only amongst the migrants

themselves, but also in meeting with the population in the host area.

30

There is nevertheless no ignoring the controversy on the subject, and there is no clear

understanding of how different kinds of presence might influence outcomes. In this

Thesis, official development aid will be investigated. The perspective of aid‟s

mediating effect, on resource scarcity in particular, is challenged by claims of relief

agencies and host government policies unknowingly fuelling conditions under which

resource use conflict may result in violence (Suliman 1999a). Counter-intuitively to

popular belief, humanitarian assistance to refugees may have exacerbating effects on

conflict (Lischer 2003: 5). Tarp and Hjertholm stress that humanitarian aid may both

intensify and prolong conflict when fallen into the hands of belligerents. The giving of

aid to combatants based in refugee camps is another problem, and a longstanding one

at that. However, it is emphasised that the unintended support of combatants is not the

fault of humanitarian aid per se, but stems from the failing security for its delivery

(2000: 306-307). The effects and effectiveness of foreign aid as a stabilizer in and

after conflict is obviously controversial, and when intended effects fail to materialise,

frustration and disillusion may follow.

[t]o be very candid…who the hell cared about Rwanda? Who really

comprehends that more people were killed, injured ad displaced in three and a

half months in Rwanda than in the whole of Yugoslavian campaign, in which we

poured sixty thousand troops and more. The whole of the Western world is

there- we’re pouring billions in there… (Gourevitch 1998 in Tarp and

Hjertholm 2000: 316).

Studies show that foreign aid at least reduces the duration of conflict, and it is claimed

that foreign aid may be an important tool for policy-makers and aid agencies alike, in

preventing future conflict (Ree and Nillesen 2006: 17). Aid can claim to be important

in limiting the intensity-level in conflict between locals and migrants. This may be

enabled through the establishment of structures created in order to reach different

groups. Organisations can enjoy greater trust among local groups, relative to the

31

current government. Suhrke argues that international relief does indeed have a

mediating effect on conflict, by tending to a migrant population‟s most immediate

needs, and thereby releasing receiving areas of this additional potential strain. This is

suggested to reduce the conflict potential in a given area (1997: 270).

As Kahl (2006), Salehyan also stresses the importance of the state, and its mitigating

effects on environmental degradation and conflict, but adds an emphasis to the

urgency of international donor assistance to countries (especially developing

countries) experiencing deteriorating environmental conditions. Given the interaction

of poverty and environmental stress, more severe outcomes are to be expected relative

to areas with better resources for crisis management. The recommended international

assistance does not necessarily stem from altruistic motives. Generous assistance

programs restrain potential and actual emigration at its source, and is considered

considerably more efficient than reactive immigration enforcement initiatives (2005:

15-16). Myers, a pioneer in the environmental refugee problematic, was sure to

emphasise foreign aid when presenting his influential predictions of potential

displacement. He stressed the importance of aid reaching the main countries and

regions of concern to the likely future production of environmental refugees. He

claimed that these regions hosted the poorest of the poor, and estimated that this would

“relieve the problem while it is still becoming a problem, i.e., before it becomes

entrenched”. Those in absolute poverty are most likely to migrate as a result of

environmental distress (1997: 178). Fundamentally however, it is of crucial

importance to suitably recognize, comprehend and respond to the issue by dealing with

the assumed sources of the problem, rather than reacting to its symptoms (ibid: 181).

The assumption in this thesis is that international aid will have a mitigating effect on

the conflict potential in a country. The data used in this analysis is based on official

development assistance (ODA) and official aid donations (per capita). One may argue

that such contributions are relatively less likely to stray, be more efficiently spent and

32

have the opportunity to finance bigger projects than private donations. This is however

not certain.

Hypothesis III: International presence reduces the risk of conflict due to

migration pressure in resource scarce areas.

33

3.0. Research Design and Data

Tying together the theoretical assumptions that underlie this thesis‟ hypotheses,

including concerns about the nature of conflicts, is a neo-Malthusian perspective on

population pressure, the relevance of resource scarcity, and an expectation that the

causal relationships between these elements are measurable and statistically

significant. One of the, perhaps, most controversial aspects of these theories is

resource scarcity, which is heavily advocated by Homer-Dixon and colleagues as

having a legitimate independent causal role:

Scholars and policymakers must take into account the independent causal role

of environmental scarcity if they are to understand and respond to many

important cases of civil violence around the world (Schwartz, Deligiannis and

Homer-Dixon 2001: 278)

The primary objective of my master thesis is, as mentioned, to investigate the

relationship between migration pressure, resource scarcity and violent conflict. By

applying a quantitative research design and using Stata software, I will analyze how

added migration pressures in areas where renewable resources are scarce affect the

probability of violent conflict (both conventional and non-conventional). Whether

international aid can mediate the risk of conflict is also examined.

Data have been collected and compiled from already existing datasets. These sources

are the Uppsala University and the Peace Research Institute Oslo (PRIO), the United

Nations Population Division, Urdal (2005), Gleditsh et al. (2002), Salehyan and

Gleditsch (2007) and the United States Committee for Refugees and Immigrants

(USCRI). The time-series is 1950-2007, the 2007-limit being determined by the

dependent variable (non-state conflict), which unfortunately is not updated further.

However, the observations are overall so many, and the time-span of such length, that

34

this will suffice for a thorough analysis. The units of analysis are limited to those used

in the dataset of the dependent variable, which are the same countries that base the

PRIO armed conflict dataset. The amount of units of analysis is considered sufficient

for the purpose of this thesis.

3.1. Previous Empirical Studies

Previous empirical (quantitative) research on the subject of migration, environment

and conflict, have produced a diversity of findings. Although there have been

conducted studies on all the main relationships presented in this thesis, the

combination factor is what justifies this particular study. The resource scarcity and

conflict nexus seems to be increasingly debated parallel to its rise on the political

agenda. Gleditsch (1998) more generally comments the literature by emphasising that

despite an abundance of declarations on the relationship between environment and

conflict, there is, as of yet, no consensus considering the causal mechanisms neither

empirically nor theoretically (ibid: 383-384)8. Homer-Dixon claims to find evidence of

noteworthy relationships between resource scarcity and conflict. Both his theoretical

and empirical findings have however been equally criticized (Gleditsh 1998; Gleditsch

& Urdal 2002; Homer-Dixon 1994; 1995; 1999; Homer-Dixon & Blitt 1998; Peluso &

Watts 2001). The notion of „conflict‟ is often operationalized as civil war. This may be

due to empirical studies finding insignificant correlation(s) between the outbreak of

war between states and resource scarcity. It is recognized that although resource

scarcity is frequently considered a theoretical source of intergovernmental conflict,

this is rarely supported empirically. A pioneer study that quantitatively investigates the

relationship between resource scarcity and the incidence of internal armed conflict is

Hauge and Ellingsen (1998), also focusing on neo-Malthusian reasoning. These find

that freshwater scarcity, land degradation, population density and deforestation

increase the risk of civil conflict. Also de Soysa (2002) has found support for a link

8 This is also highlighted in Tir and Diehl (1998), where the literature suggests a link between

population variables and international conflict (interstate conflict).

35

between high population density and internal armed conflict using large cross-national

time-series studies. However, empirical research has, so far, failed to find unanimous

evidence for a resource scarcity and conflict scenario. Even in an attempt to replicate

the study of Hauge and Ellingsen (1998), Theisen (2006) did not find the empirical

evidence to support the claims of the original authors.

Despite population growth and density being associated with increased risk, Raleigh

and Urdal (2007) find that the effects of political and economic factors far outweigh

those between demographic and environmental factors at the local level, and conflict.

The effects of land degradation and water scarcity are weak or insignificant. Claims of

a „new era of insecurity‟ following the end of the Cold War, is by Urdal (2005) found

to have little empirical support, and the suppositions of environmental, and

demographic factors threatening security and state stability are thereby unfounded.

Boserup‟s (1965) general findings rather points to population pressure as a prerequisite

for modernization as it gives rise to economies of scale, and in a later study finds that

scarcity of arable land at an aggregate level appeared to reduce the risk of conflict

(1981). Myers (1992) voices the notion of steep population growth and a

discontinuous process of environmental change. This entails the degradation of a given

resource exploited at a given level, until the exploitation exhausts the self-renewable

capacity of that resource (Myers 1992: 116).

Influxes of refugees into an area can cause considerable stress on natural resources,

leading to both environmental and social impacts (UNHCR 1996; Black & Sessay

1997) Focusing on environmentally induced migration (environmental refugees) a

major and interesting conclusion in Suhrke‟s study is that insofar as environmental

degradation causes displacement of people, it is more likely to generate exploitation

rather than conflict (1993: 35). Salehyan and Gleditsch quantitatively link refugee

migration to civil conflict in receiving states (2006). There are presented several

interpretations of this. Economic and resource competition, disruption of a previous

36

ethnic balance, and the spreading of arms and ideologies all hold explanational

powers. Hugo emphasises the international dimension of the migration-environment

relationship, which he assumes to be of increasing significance. He investigates the

relationship between migration and the environment, and calls for increased

international involvement (1996). Raleigh, Jordan and Salehyan consider the impact

environmental change has on migration and conflict. They argue that changes in

migration patterns due to the environment, is highly unlikely. They further claim that

environmental migration tends to be short term and internal, and the potential for

instigating conflict is low (2008). Weiner and Teitelbaum (2001) argue that it is

appropriate to suggest a connection between demography and levels of violence, given

that that worldwide increase in displaced populations such as refugee flows and

internally displaced persons have been among the most notable developments over the

past quarter century (ibid: 51).

3.2. Method: Logistic Regression

Given that the dependent variable is dichotomous with the values conflict (1) and non-

conflict (0), the most suitable methodological design is a maximum likelihood logistic

model. The dependent variable is two-fold, entailing individual analysis of both

conventional conflict and non-state conflict. The relationships between the dependent

and the independent variables are not expected to be linear; as such a regression model

would suppose possibilities for predicting values of Y outside the 0-1 interval. The aim

of the analysis is to investigate how the share of having a particular value on the

dependent variable varies for different values on the independent variables. Shares can

by definition only vary between 0 and 1, so answering to this; these are transformed to

log odds of having a particular value on the dependent variable. The logarithmic

transformation extends the range of the dependent variable from - ∞ to + ∞ (Skog

2007: 351-366). The challenge lies in finding (co)variations between the units, and

37

find significant correlations between the dependent and independent variables. A

logistic model can be written as (ibid: 358):

To facilitate interpretation of correlation between variables, odds ratio, rather than log

odds are applied9. Odds ratios are the antilogarithms of the log odds (the regression

coefficients). Odds ratio measures the relative change in the odds of having the value 1

in the dependent variable resulting from a one unit increase in an explanatory variable,

controlled for all other variables (Skog 2007: 364-365).

3.3. Time-Series Regression Analysis

Time-series analysis allows studying successive variations in the dependent variable at

one or several observation units over longer periods of time. This method can help

determine the nature of the effects the independent variables have on the dependent,

and may determine whether these are durable or temporary phenomenon‟s, and

whether they emerge gradually or abruptly (Skog 2007: 82). A commonly used

definition of this type of analysis argues that “time series analysis accounts for the fact

that data points taken over time may have an internal structure (such as

autocorrelation, trend or seasonal variation) that should be accounted for”

(NIST/SEMATECH). Time-series analysis includes a spectrum of exploratory and

hypothesis testing methods that have two main goals. The first goal involves

identifying the nature of the phenomenon represented by the sequence of observation.

The second goal represents an ability to forecast (predicting future values of the

9 Implies the use of the ‟logistic‟ rather than ‟logit‟ command in the statistical tool Stata.

38

variable), based on the previous time-series. Both of these goals assume an

identification of pattern(s), and a formal description of these (StatSoft)10

.

3.4. Operationalization of variables

3.3.1. Dependent Variable

3.3.1.1. Violent Conflict

The dependent variables are based on two Uppsala/PRIO conflict datasets entailing

both conventional and non-state conflict. For these two „Violent Conflict‟ variables

there is a lowered intensity requirement of 25 battle-related deaths per year, rather than

the traditional limit of 1000. Conventional conflict data presupposes “a contested

incompatibility that concerns government and/or territory where the use of armed

force11

between two parties, of which at least one is the government of a state”

(Gleditsch et al. 2001: 619). Non-state conflict understands the 25 battle-related death

minimum as results of “the use of armed force between two organized armed groups,

neither of which is the government of a state, which results in at least 25 battle-related

deaths in a year” (UCDP 2009). The lowered battle-related death criteria is not only

due to the data alone, but is very much motivated by theoretical arguments that imply

that the kind of conflict to be expected in a resource scarce setting is that of low-

intensity. These data reflect conventional, as well as non-state conflict. The last

10

This type of data is often complimented by cross-sectional analysis in order to reduce the

risk of spurious correlations due to underlying factors that represent problems in the analysis

of some variables, is not present, or at least less present with other units of observation (Skog

2007: 83). This is also done in this analysis, although the comments on this follow in

footnotes, as it is considered but a valuable supplement.

11

The „use of armed force‟ may need further specification and refers to the “use of arms in

order to promote the parties‟ general position in the conflict resulting in deaths (the

UCDP/PRIO Armed Conflict Dataset Codebook Version 4-2009: 1).” „Arms‟ concern “any

material means, e.g. manufactured weapons but also sticks, stones, fire, water etc (ibid)”.

39

mentioned type of conflict may in itself allow a more thorough registration of conflict

relevant to the principal independent variable of renewable resource scarcity.

The conventional conflicts available by UCDP/PRIO use the common civil war

operationalization-procedure for coding „onset‟. This entails coding every outbreak of

conflict either as a result of entirely new conflicts, or if two or more years have passed

since casualties were below the required level. When considering „onset‟, the different

conflicts within a country will not be taken into account, and the outbreak of a conflict,

be they related or not, always needs to meet the 2-year requirement of fewer than 25

battle-related deaths. If there are observed several outbreaks of conflict in the same

year, the first actual outbreak will set the basis for the evaluation of the 2-year “peace”

criteria. The dependent variable conflict onset, is coded 1 for the first year of a conflict

and 0 if no conflict takes place in the state in that particular year (dichotomous).

Subsequent ongoing years of the same conflict are dropped from the estimation

sample. In cases where there are multiple conflict onsets in a country, data on a new

onset is included if it occurred during the years when another conflict was ongoing

(Salehyan & Gleditsch 2006)12

. Reliable data below a two-year limit of inactivity is

considered difficult, although there is no authoritative answer to how conflict should

be coded (Strand 2006: 9).

The dependent variable indicating non-state conflicts were initially thought to also

measure onset. However, given that the data does not specify the end of a conflict and

that there are only available data for five years, this is difficult13

. Therefore, non-state

conflicts are coded by incidence. The amount of conflicts in a given year will not be

12

Alternative measures of coding new „onset‟ cases when new rebel groups enter the conflict,

have been estimated by Salehyan & Gleditsch, resulting in insignificant variations of results

(2006: 350 n. 45).

13

This information, together with other updates, will be available in the updated version of

this dataset due in August of this year. The data will then also cover the years from 1989

throughout 2009

40

taken into account. The variable is coded 1 in the case of conflict, and otherwise holds

the value of 0. These data are limited to sufficiently few years, so that issues of time-

dependency are limited. In addition, countries with conflict in the first year of

registered data (2002) would be coded as „onset‟ while it is probable to assume that all

of these did not break out then.

3.4.2. Independent Variables

3.4.2.1. Migration Pressure

Migration is a complex phenomenon, and thereby poses difficulties for reasonable

operationalizations of the concept. Aware of these challenges however, and the

necessary limitations following the operationalization, I am fairly confident in having

found a sufficiently explanatory measure. This study is not concerned with a state‟s

population per se, but rather the additional demographic pressure as a factor external to

natural population growth. Variations of the proportions of this additional pressure

over time, and investigated relative to the availability of arable land is what is of

interest, as it is the scarcity of resources that is hypothesised as paramount to the

likelihood of the outbreak of conflict. Migration pressure is hereby operationalized as

forced migration, and is defined by the United Nations High Commissioner for

Refugees (UNHCR) as:

[A person] owing to well-founded fear of being persecuted for reasons of race,

religion, nationality, membership of a particular social group or political

opinion, is outside the country of his nationality and is unable or, owing to such

fear, is unwilling to avail himself of the protection of that country; or who, not

having a nationality and being outside the country of his former habitual

residence as a result of such events, is unable or, owing to such fear, is

41

unwilling to return to it (Article 1, The 1951 Convention Relating to the Status

of Refugees).

The focus in this thesis is forced migration crossing borders, namely refugees,

however, their reasons for fleeing being of less importance. The refugee data used in

this thesis stems from two different sources, as the time-series vary, and these will be

investigated separately. Salehyan and Gleditsch‟s ”Replication data for refugees and

the spread of civil war” (2007) and the Centre for Systemic Peace‟s (CSP) data on

”forcibly displaced populations, 1964-2008” (USCRI, annual). The former dataset

have gathered information from the Population Data Unit of the UN and the High

Commission for Refugees (UNHCR) and cover the years 1946 until 2003 (the time-

series of this analysis begins in 1950). The refugee data made available by CSP and

compiled by USCRI, spans from 1964 until 2008 (the upper limit for this analysis‟

time series being 2007). The information used from these datasets concern the number

of incoming refugees to a host country, and these datasets are by far the most complete

on the issue. All refugee data are log transformed in order to reduce the impact of

extreme values.

3.4.2.2. Renewable Resource Scarcity

Renewable resources are natural resources that theoretically regenerate themselves

indefinitely, through normal ecological processes. Scarcity of such resources emerge

when the flow or stock of the resource is quantitatively exhaust or qualitatively

degraded at a rate faster than its regeneration, or distributed so as to artificially deprive

individuals of the resource (Kahl 2006: 31). Population growth, be that on a national or

international scale, has often been the focal point of models of environmental

degradation. Growing populations have been considered one of the factors that greatly

exacerbate poverty, starvation, economic stagnation and resource depletion (Tir &

Diehl 2001: 58). Homer-Dixon emphasises that what should be investigated are not

42

absolute sizes but rather resource supply relative to the demand and social distribution

of that resource (Schwartz, Deligiannis & Homer-Dixon 2001: 275).

Common forms of environmental degradation include land degradation,

desertification, rising sea levels induced by global warming and deforestation with its

many consequences (Suhrke 1993: 3). This thesis primarily focuses on arable land and

its availability.14

Theoretically this focus is a natural limitation. Homer-Dixon, as a

chief provider of premise concerning resource scarcity and conflict, defines renewable

resources as being cropland, water, forest and fisheries (Homer-Dixon & Blitt 1998).

The most important of these resources are cropland and freshwater (Homer-Dixon

1999; Kahl 2006). Homer-Dixon believes that environmental scarcity has “it‟s most

profound effects on people‟ lives in rural areas” (Homer-Dixon 1999: 166). He claims

that land stress and bulging populations may generate waves of refugees that spill

across borders, having destabilizing effects on the recipient‟s domestic order and on

international stability (Homer-Dixon 1991: 77). Myers (1992) argues that population

growth in conjunction with environmental degradation that to a great extent stems

from this very population growth, overwhelms the environmental underpinnings of

agriculture. A state‟s functional capacity heavily relies on the agricultural sector‟s self-

sufficiency in times of demographic and environmental stress (Kahl 2006: 47). It may

also be argued that this resource indirectly takes into account a certain freshwater

availability, since this a predicament of agriculture. Brown follows this reasoning by

arguing that agricultural commodities may very well function as proxy for key natural

resources, like water and land, and can increase the risk of competition, and perhaps

conflict, over scarce resources (2008: 1).

14

There are several alternative measures. Other frequently mentioned resources are fisheries

and forests. Arable land, however, seems to be most fitting in this context. First and foremost,

the focus on arable land can be legitimized theoretically. Secondly there are significant data

challenges both concerning availability and the diverting understandings of the mere

measures of the mentioned alternative resources.

43

To illustrate this logic statistically, the „Potential Cropland‟ (per capita cropland

scarcity) variable will be an important operationalization of the renewable resource

scarcity concept. The operationalization demonstrates renewable resource scarcity by

creating a variable of productive soil relative to population size, further indicating

population pressure on these resources. The variable is defined as “all of a country‟s

land that falls into the following land use categories: arable land, permanent crops,

permanent pastures, and forests and woodland” (Urdal 2005: 424). This variable is log

transformed given that there is reason to believe that there is no linear relationship

between density and conflict15

(Tir & Diehl 2001). This transformation also excludes

extreme cases possibly disturbing the results16

. This measure originates from Urdal‟s

study of 2005, and therefore needed a seven year update considering the scope of this

analysis. The „Potential Cropland‟ variable represents the population pressure on the

amount of land available for agriculture. Assuming insignificant changes in arable land

during recent years, this variable was updated by calculating the annually updated

population of each country relative to their respective year 2000 value on arable land17

.

The source data for these calculations are developed by the UN Population Division

(Demographic Yearbook, annual), the World Development Indicators (World Bank),

World Factbook (CIA annual), Encyclopaedia Britannica (Britannica annual) and The

Statistical Abstract of the World (Reddy, 1994) for small states.

The independent variable mapping the share of rural versus urban population within a

given country is meant to give an idea of the domestic context. There is little and

incomplete data on agricultural populations, resulting in this alternative measure. The

assumption is that the rural population primarily relies on agriculture, while the urban

15

When a country reaches a certain level of density, it is probable to expect a decrease in

effect. It is a larger transition for a smaller country to have a population density go from 200

to 1200, than it is for a larger country to go from 5000 to 6000

16

Extreme-value countries such as e.g. Singapore and Bahrain.

17

This is also the assumption in the original study, where the potential cropland is based on a

single observation per country only dating from the 1993-2001 period (Ural 2005: 424).

44

population does not. This is a more inaccurate, but arguably reasonable, proxy. The

information used directly in the analysis is the percentage of rural population. These

data stem from the UN Population Division (UN 2008), and spans throughout the

time-series of the analysis (1950-2007) with data updates for each country every fifth

year.

3.4.2.3. International Aid

International aid includes net official development assistance and official aid

(measured in current US $). A limitation of this variable is that it lacks a geographic

component, which can mean that in practice, the aid resources can be used in different

parts of the country other than those areas hosting refugees. Nonetheless, there are

other consequences of receiving aid that may be considered to ameliorate a refugee

situation, and thereby reduce the potential conflict level. There is a direct effect given

that aid presupposes a system of distribution that in turn can enable a more efficient

handling of refugees, compared to countries where this is non-present. Foreign aid can

also serve as an indicator of the degree to which donor-countries would be interested,

or likely, to get involved in a crisis- or refugee situation. By this logic, high levels of

foreign aid should indicate a willingness of third parties to contribute, if the situation

so requires. To best measure this presence, data on the amount of aid per capita will be

used. These data are collected and developed by the World Development Indicators

(World Bank, 2008).

3.4.3. Control Variables

There are two different development indicators in the analysis. The Infant Mortality

Rate (IMR) is expected to better fulfil its intention, but the traditional measure using

45

Gross Domestic Product (GDP) per capita (Gleditsch 2002)18

will serve as a control, as

well as comparison. Using country-year as the unit of analysis, and the onset of

conflict as the dependent variable, problems of autocorrelation may occur. A country

having experienced conflict in a given year is inherently likely to be in conflict the

following year also (Hauge & Ellingsen 2001: 47; Hegre et al. 2001). In order to

reduce the potential impact of this problem, the „brevity of peace‟ variable is used to

control for conflict in the previous year/ years since last conflict (Hegre et al. 2001).

This variable controls for time dependency in the study of conflict, and assumes a

positive relationship between past conflict and the risk of new conflict onset. The

rationale predicts the temporal dependence to be strongest directly after a conflict, and

thereon decrease with time. It may be expected that the correlation between population

pressure and conflict will decrease after reaching some threshold. “States can handle

only so much conflict and at some limiting point additional population pressure will

not produce any greater likelihood of conflict involvement or escalation (Tir & Diehl

2001: 69).” By this logic the total population data is log transformed.

Regime is theoretically argued to integrate the variables of opportunity structure and

state capacity (Schwartz, Deligiannis & Homer-Dixon 2001: 279). The Polity IV

scheme examines “concomitant qualities of democratic and autocratic authority in

governing institutions”. The "Polity Score" illustrates this regime authority spectrum

on a 21-point scale ranging from -10 (hereditary monarchy) to +10 (consolidated

democracy) (Marshall & Jaggers 2002)19

. The effect of democracy on conflict is

debated, and the very term „democracy‟ is claimed to be used too loosely by lay

commentators and experts alike. A comprehensive understanding of the term includes

an extraordinarily complex “set of social phenomena and institutions that have

complicated and multiple effects on the incidence of social turmoil and violence

(Homer-Dixon 1999: 182)”. Theories concerned with democracy and conflict tend to

18

These are updated until 2004, while the data on IMR stretches throughout the period of

analysis (1950-2007). 19

The Polity variable is imputed for all years that miss or that uses transition codes. The

„polmiss‟ variable is intended to discover the possible distortion that this may provoke. This

dummy variable is included to avoid loosing too many observations.

46

indicate that democracies experience less conflict and rebellion, relative to autocracies

and in-between regime types. However, note that the democratization process itself

might generate conflict (Hauge & Ellingsen 2001: 42). Others claim that it is

democracies in transition, meaning regimes in a process of change, are most inclined

to experience conflict. This suggests that the democratization process in itself may

generate conflict (Hegre et al. 2001). If the state failure hypothesis is an important

pathway to armed conflict, it should be expected that statistical controls for low state

capacity and state failure should capture some of the explanatory power of the

demographic and environmental variables. Previous studies have shown curvilinear

coherence regarding the Polity variable, so the quadratic value of this variable is

introduced in order to capture this effect. The control variable Polity squared checks

for curvlinearity. The variable thereby assumes that the middle-values are those most

prone to affect the dependent variable. This is theoretically defended by the

assumption that the societies in polity transition are more likely to experience conflict,

in contrast to the extreme values on this scale, that are either perfectly authoritarian or

democratic (Hegre et al. 2001).

3.4.4. Validity and Reliability

Determining the quality of the analysis is influenced by validity and reliability

assessments. Reliability is concerned with the degree to which data is measured

without error, and implies a notion of replicability (Theisen 2006: 62). This entails the

precision of measuring data. The validity depends on what is measured, and whether

these entail the qualities and characteristics of the relevant research question. It

indicates the quality of the empirical operationalizations of a theoretic definition

(Hellevik 2002: 52-53, 471). The validity also suggests whether the indicators used

actually fit the causal arguments of the theory. The reliability of the data is critical for

the validity of an analysis. High reliability is a necessary, but not sufficient, condition

47

for high validity (Theisen 2006: 62). To ensure a defensible level of validity, necessary

assumptions of logistic regression must be satisfied. The different data used for the

analysis in this thesis have, almost all of them already been used in published studies

(Urdal 2005, Salehyan & Gleditsch 2006), and stem from recognized sources of data

collection (World Bank, United Nation Population Division etc). The reliability of the

data is expected to be quite high. There are many observations, and no alarming

amounts of missing20

. Given that one of the main explanatory variables (refugees) has

two different sources of data, the reliability should increase further, ensuring more

accurate measures (this also applies to the development indicator IMR and GDP).

The operationalizations of „migration pressure‟ and „renewable resource scarcity‟ are

perhaps the variables that deserve the most scrutiny in regard to validity. The

requirement of quantifiable values was a main concern when operationalizing

migration pressure. Most migrate within their state‟s borders, and are so called internal

migrant or internal refugees. Accurate numbers of these are however difficult to find,

although there is an increase of interest concerning such migration patterns.

Additionally there is no consensus on their very definition, making it difficult for

empirical studies to determine who falls under this category. There are reasons to

believe that internal migration poses less additional strains on resources relative to

external migrant or refugees, indicating that cross-border migration is more relevant to

the dependent variable (violent conflict).

Renewable resource scarcity has not such an obvious solution as had the migration

variable. When it comes to renewable resource scarcity, there are challenges both due

to data and the conceptualization. Inspired by Urdal‟s solution to this problem,

potential arable land, having already been theoretically identified as a main resource,

together with population forms the „potential cropland‟ variable (2005). This

20

Only the „aid‟ variable has substantial amount of missing, which is taken into account

throughout the Thesis.

48

operationalization includes a population pressure aspect by considering scarcity

relative to the population in the respective states, further increasing its relevance to the

theoretic material of the thesis. The controls used, are all „standard‟ in quantitative

studies on civil war.

49

4.0. Time-Series Analysis21

The three previously presented hypotheses of this thesis expect that migration pressure

(operationalized as refugees) and constraint on renewable resources (per capita

availability of arable land) are problematic, and thereby increase the probability of

conflict (both conventional and non-state). Aid (per capita) is further assumed to

mediate the effects of migration. These assumptions are statistically explored through

logistic regression analysis. The countries used in this analysis are restricted to those

that enjoy independence, and meet the minimum requirement of 250 000 inhabitants

(in year 2000)22

. Migration pressure, operationalized as refugees, is the independent

variable with the most emphasis. Two different sources of refugee data are used to

ensure the most correct results as possible. The analysis, as a whole, spans over the

years 1950 to 2007. The refugee data gathered by Salehyan and Gleditsch (2007) cover

the years 1950-2001, and the data provided by the Center for Systemic Peace, last

from 1964 to 2007. These data are analysed separately, and combined constitute the

data material on refugees allowing full coverage of the time-period of the analysis. In

all models, the main explanatory variable, indicating migration pressure, is examined

alone together with the control variables. In order to more closely examine the data

and either confirm or deny the presented hypothesis, the models will subsequently be

expanded by the inclusion of the other variables and interaction terms as well as the

controls.

The Malthusian perspective is integrated in the analysis by the very variables under

scrutiny. Although the theory is complex, the investigation of the above-mentioned

variables is a reasonable simplification of main concepts. Likewise, it is difficult to

21

All models have been run using both IMR and GDP separately, as development indicators.

Although it does not explain the amount of variance, the Pseudo R2 indicates that IMR is

slightly better suited for the majority of the models. The IMR measure will therefore be used

when presenting the different models.

22

The country list is, throughout the analysis, based on an updated version of the dataset of

Urdal 2005.

50

capture all aspects of Homer-Dixon‟s understanding of environmental scarcity.

„Demand-induced‟ scarcity is measured by population growth, and further

operationalized as incoming refugees. „Supply-induced‟ scarcity is measured as land

degradation, and operationalized as per capita availability of actual and potential

cultivable land. The politics of distribution entailed in the notion of „structural

scarcity‟ is an effect not easily captured, at least not in this analysis, nor is it a priority

of the thesis. Kahl sums up the reasoning of Goldstone and Homer-Dixon, which

clarifies the expectations of the following analysis and the emphasis put on different

aspect of the theory (population and environmental pressure, agriculture, aid and state

capacity), resulting in the before-mentioned variables.

[...] population and environmental pressures in developing countries often

generate intense hardship among agricultural labourers and the urban poor.

They contend, however, that strong capable states are typically able to prevent

such deprivation coalescing into organized violence through a mix of relief for

aggrieved individuals, co-optation of opposition leaders, and outright coercion.

Therefore, large-scale violence is only likely to occur when social grievances

emanating from rapid population growth, environmental degradation, and

natural resource scarcity combine with eroding state authority and escalating

intra-elite competition (Kahl 2006: 10).

My analysis is not limited to the developing world, and has undergone necessary

simplifications in order to conduct analysis.

51

4.1. Conventional Conflict

When only considering the effects of migration pressure (refugees) in the models of

conventional conflict (using both sources of refugee data), these are highly significant

and thereby in accordance to the findings of Salehyan and Gleditsch (2006). This is

reassuring, and provides an excellent basis for the remaining analysis. These authors

understand the effects to imply a spreading of conflict by refugees, entailing the import

of ideology arms etc. If the theoretical assumptions presented in this particular thesis

are correct however, these effects should at least be partly dependent on high

renewable resource scarcity and absence of international aid (and to a certain extent in

a context of large rural populations).

Table 1: Conventional Conflict Logistic Regression Results, 1964-2007 (Odds Ratio)

Model Model Model Model Model Explanatory Variables 1.0.a 1.0.b 1,1 1,2 1,3

Refugees_CSP 1.051*** 1.048*** 1.056*** 1.046*** 1.048*** Potential Cropland 0,968 0,971 0,947 0,968

Aid per capita 0,998 0,998 0,998 0,998 Rural Population 1,007 1,006 1,007 1,007 Interaction Terms

Refugees & Rural Population 0,999

Refugees & Potential Cropland 0,989

Refugee & Aid 1,000 Control Variables

Infant mortality rate 1.010*** 1.007*** 1.007*** 1.008*** 1.007*** Total Population 1.261*** 1.358*** 1.256*** 1.367*** 1.258*** Regime 1,008 1,005 1,005 1,005 1,005 Regime, squared 0.985*** 0.986*** 0.986*** 0.986*** 0.986*** Missing regime data 1,649 1,040 0,92 1,147 1,050

Controls for statistical dependency

Brevity of peace 2.746*** 3.005*** 2.981*** 2.971*** 3.000*** N 4927 4586 4586 4586 4586 Log likelihood -613,021 -576,380 -575,961 -575,913 -576,369 Pseudo R2 0,096 0,095 0,095 0,095 0,095

52

NOTES: * p < 0.10, ** p < 0.05, *** p < 0.01.

Italics = part of interaction term.

The migration pressure indicators remain significant at a level of one percent, even

with the inclusion of the other variables, and in respect to both sources of migration

data. There are minimal changes in the Odds Ratio. None of the newly added variables

reach significance, and both migration pressure and most of the control variables are

consistently significant through all models. Neither significance nor effects seem to be

altered to any large extent.

Table 2: Conventional Conflict Logistic Regression Results, 1950-2001 (Odds Ratio)

Model Model Model Model Model Explanatory Variables 2.0.a 2.0.b 2.1 2.2 2.3

Refugees_S&G 1.052*** 1.044*** 1.055*** 1.043** 1.045** Potential Cropland (log) 0,981 0,983 0,984 0,981

Aid per capita 1,000 1,000 1,000 1,000 Rural Population 0,999 1,002 0,999 0,999 Interaction Terms

Refugees & Rural Population 0.999*

Refugees & Potential Cropland 0,997

Refugee & Aid 1,000 Control Variables

Infant mortality rate 1.006*** 1.007*** 1.007*** 1.007*** 1.007*** Total Population (log) 1.214*** 1.278*** 1.281*** 1.281*** 1.279*** Regime 1,004 1,004 1,002 1,004 1,003 Regime, squared 0.985*** 0.985*** 0.986*** 0.985*** 0.986*** Missing regime data 0.102** 0,278 0,297 0,279 0,284

Controls for statistical dependency

Brevity of peace 3.465*** 3.299*** 3.250*** 3.311*** 3.300*** N 5921 4606 4606 4606 4606 Log likelihood -718,947 -596,700 -595,282 -596,667 -596,667 Pseudo R2 0,088 0,088 0,090 0,088 0,088

53

NOTES: * p < 0.10, ** p < 0.05, *** p < 0.01.

Italics = part of interaction term.

Including interaction into these models has little implication23

. When using the more

recent CSP data the variables stay very much unchanged, with none of the interactions

being even remotely significant. The migration pressure indicator remains significant

at a one percent level throughout, and the control variables also keep their

significance, as well as their effects. The Salehyan and Gleditsch data, being slightly

more dated, again shows mainly the same results. The refugee variable stays

significant throughout all models, as do the control variables. There are small changes

in Odds Ratio and no interactions are highly significant. However, the interaction term

of refugees and rural population is significant at ten percent, with negative effect. This

could imply that the impact of added migration pressure on conflict potential is higher

in countries with lower rural population, which in turn can be interpreted as countries

with higher population density. This reasoning fits with the neo-Malthusian logic of

population pressure‟s impact on conflict. A fixed population pressure is expected to be

perceived relatively more intense in areas already experiencing a higher population

density. This negative effect, however, seems to be weak. Considering the Pseudo R2,

there is very slight reduction in the models when adding more variables no matter the

migration data used. Although difficult to interpret substantially, comparing the

Pseudo R2 values of the different models may be useful as the increase in value

indicates a decrease in log likelihood, implying an improved model24

.

23

Although no main focus of this thesis, both conventional conflict and non-state conflict

tested for the possible interaction between renewable resource scarcity and rural population.

This could indicate whether resource scarcity is of greater or less importance to the risk of

conflict in a country, relative to its size of rural populations. None of these interaction terms,

however, proved significant.

24

Also, however difficult to interpret, given the amount of observations in the analysis the

Prob>chi2 at least implies that the models cannot be rejected.

54

As has been shown in regards to conventional conflict; migration pressure is

significant, unlike any of the other independent variables. The significance of the

migration pressure impact on conventional conflict onset, could lend support to the

perspective of conflict „contagion‟ effects, which previously has found empirical

support (Salehyan & Gleditsch 2006, Forsberg 2009). The contagion hypothesis

associates refugee flows with destabilising spill-over effects, which indicates that host

countries receiving refugees from countries in civil war, are more prone to conflict

than states not receiving such refugees (Forsberg 2009b: 20). These studies connect

refugees and conflict onset empirically, such as is done in this thesis, but to thereby

interpret this as support for this theory, will be a little too hasty. The results of

Salehyan and Gleditsch in their study of “Refugees and the Spread of Civil War”

(2006), find empirical evidence suggesting that refugee flows represent a driving force

in the diffusion of civil conflict, and that this in turn creates security concerns for host

countries. They emphasise, however, that this constitutes no deterministic links

between refugees and conflict, and that the majority of cases of refugee flows do not

result in violence (ibid: 360-361).

Comparing coefficients between our studies, mine have stronger effects on conflict

onset (0.042 versus 0.051)25

when using the same refugee data and only control

variables26

. When including all other contextual variables this difference in effect is

lost (0.042 versus 0.043)27

(Salehyan & Gleditsch 2006: 355). Salehyan and Gleditsch

investigate the likelihood of refugees from neighbouring countries influencing the risk

of conventional conflict onset. The refugees accounted for are thereby those that

migrate from areas close by. One should think, given the reasoning of their study, that

looking at this group isolated would result in a clearer impact on conflict. The refugees

25

27

See Appendix 7.3.

26

Given the approximate number of observations, and using the same refugee and conflict

data, despite being difficult to interpret, the coefficients may be compared to certain, though

limited, extent.

55

are assumed to have more interest in mobilizing in favour of a warring party when it is

considered that this may have plausible effects in their country of origin (See also

Wiener & Teitelbaum 2001; Lee 2001). This, however, does not seem to be the case as

my investigation of refugees irrespective of „area-attachment‟ have similar effects on

conflict onset, if not even stronger effects. It would be presumptuous to suggest denial

of the “contagion effect”, but it would, perhaps, encourage further scrutiny28

.

4.2. Non-State Conflict

It is worth remembering that the non-state data have a considerably shorter time-series

compared to the other dependent variable, conventional conflict onset (2002-2007

versus 1950-2007). When investigating non-state conflict, only the CSP refugee data

are used, given its time-series. The migration pressure variable, which is strongly

significant when it comes to conventional conflict, is insignificant concerning non-

state conflict29

. The migration effect on non-state conflict using only control variables

is also insignificant. This diverts from the (expected) outcomes of the conventional

conflict models and theoretical assumptions, such as those derived from Malthus

especially. Significance may perhaps not be reached because these data include a far

shorter time series and, but this is difficult to determine. The potential cropland

variable, however, indicating the population pressure on arable land, and thereby

28

They present in their study significant findings with stronger coefficients, but these use the

traditional civil war measure of 1000 battle deaths, and are therefore unsuited for comparison.

The authors also argue that this threshold is a problematic one.

29

When using the cross-sectional design on the non-state conflict data, the refugee variable

turns significant at a five percent level with clearly positive effects, when only including

control variables. This is in accordance to the theoretical expectations of this thesis. This

could suggest that the previous insignificant findings concerning non-state conflict and

migration pressure could be a result of an insufficient time-series, rather than uncorrelated

theory and empirical findings.

56

resource scarcity, is significant almost at the one percent level30

. Counter-intuitively,

this variable appears to have negative effect on non-state conflict, entailing a reduced

risk of non-state conflict, given higher values of renewable resource scarcity31

. These

models cannot be rejected, given the Pseudo R2 in the test of goodness-of-fit that

indicates an improvement in model fit when adding contextual variables32

.

Table 3: Non-State Conflict Logistic Regression Results, 2002-2007 (Odds Ratio)

Model Model Model Model Model Explanatory Variables 3.0.a 3.0.b 3.1 3.2 3.3

Refugees_CSP 0,962 0,969 0,947 0,973 0,990 Potential Cropland (log) 0.660** 0.652** 0.672** 0.657** Aid per capita 0,990 0,989 0,990 0.984* Rural Population 1.023* 1.024* 1.023* 1.027** Interaction Terms

Refugees & Rural Population 1,003

Refugees & Potential Cropland 0,984

Refugee & Aid 0.996* Control Variables

Infant mortality rate 1.026*** 1.019*** 1.020*** 1.020*** 1.018*** Total Population (log) 2.255*** 2.536*** 2.540*** 2.554*** 2.520*** Regime 1,023 1.069* 1.064* 1.068* 1.079*

Regime, squared 0.985* 0.976*** 0.976*** 0.976*** 0974*** Missing regime data 0,302 - - - - Controls for statistical dependency

Brevity of peace 4,920 4,178 3.767*** 4.098*** 3.499*** N 925 845 845 845 845 Log likelihood -143,012 -122,485 -121,633 -122,380 -120,672 Pseudo R2 0,340 0,373 0,377 0,374 0,382

30

When applying GDP per capita as development indicator, the Potential Cropland variable

remains significant, at a five percent level.

31

Due to amounts of missing, the independent variable „aid‟ was attempted excluded from the

model. Following this the potential cropland variable loses its significance. When testing for

interaction between the two variables, this is nonexistent. The exclusion of the aid resulted in

an increase from 845 observations to 919.

32

The considerably higher values of Pseudo R2 compared to conventional conflict is not

uncommon given both the time-series of the data and measure of onset versus incidence of

conflict.

57

NOTES: * p < 0.10, ** p < 0.05, *** p < 0.01.

Italics = part of interaction term.

The interaction between the refugee variable and rural population is insignificant, both

the control variables and the potential cropland variable remain significant. Although

resulting in minimal changes in its effect on the dependent, the resource scarcity

indicator even improves its level of significance slightly. This variable also interacts

insignificantly with refugees. The interaction between refugees and aid is almost

significant (0.07), but results in little change otherwise in the model. None of the

interactions are significant, but it is interesting to note that the potential cropland

variable remains significant through all tests. The percentage of rural population is

either significant, or close to significant, in all models. These two variables do not

significantly interact. The aid variable stays insignificant through all tests. The control

variable indicating the brevity of peace continues to be significant with relatively

strong effects.

The theoretical implications of the findings of this analysis seem to question the

presumptions of both Malthus and the relative deprivation aspects of renewable

resource scarcity literature. Neither migration nor resource scarcity have shown

expected effects. First of all, migration has no significant impact on conflict. Although

the added population pressure that refugees represent have seemingly no importance

concerning this type of conflict, the total population variable remains significant at one

percent through all models, with positive effect33

. This last variable could, however,

indicate the size of a country, rather than its population density, and is thereby no

accurate indicator. Secondly, resource scarcity indicates a reduction in the risk of

conflict. As the Potential Cropland variable is negatively significant, one should

33

This variable is also significant in relation to conventional conflict, although the Odds Ratio

effects in these cases are positive but weak.

58

expect that the rural population indicator is complementary, and thereby has the same

direction of results. This latter variable, however, has sporadically significant but

positive effects on non-state conflict, entailing that high values on this variable

indicates heightened risk of conflict.

From these results one can argue that if the total population indicator indeed measures

population density to a reasonable extent, then the considerably stronger effects this

variable has on non-state conflict compared to conventional conflict, may assume that

in densely populated areas, the likelihood of a strong central power is low and thereby

suggesting low government involvement. The strong negative effect of renewable

resource scarcity further implies that there in such areas, naturally, is no scarcity of

land, both arable (as indicated in the Potential Cropland variable) and otherwise.

4.3. Analysis Summary34

The time-series logistic analysis of conventional conflict stays relatively unchanged,

despite the inclusion of additional contextual factors, as well as when testing for

interaction. The results indicate a clear significance of additional migration pressure

(refugees) at a constant one percent level, with the approximate positive Odds Ratio

effect of 1.04. These effects persist seemingly unaffected by additional variables and

interaction terms. The control variables also remain significant. Diversions in the

34

Tests using cross-sectional analysis have been done in an attempt to give a closer look at the

possible relationship between the hypothesized variables. For the Salehyan and Gleditsch

refugee data this is year 2001, and for the CSP data (in relation to both conventional and non-

state conflict) this is year 2007. The aid variable is unfortunately excluded, as the large

amounts of missing distort the results. Non-state conflict has no significant variables or

interaction. The migration variable is significant at a 5% level with clearly positive effect, but

loses its significance at the inclusion of other contextual variables. When investigating

conventional conflict, none of the independent variables are significant, except resource

scarcity, which indicates significance at a 10% level (0.089) with strong negative effects on

the dependant (using CSP data). There is no interaction.

59

statistical findings given the use of refugee data may either be due to the accuracy of

observation or the somewhat different time-series, although it is difficult to distinguish

these effects.

Non-state conflict loses the significance of migration pressure, but indicates strong

significance regarding renewable resource scarcity at an almost one percent level. The

effects on conflict, however, are quite strongly negative. This variable remains

significant throughout all the models. It is argued frequently that the larger the role of

scarcities, the smaller the conflict (Homer-Dixon 1999; Suliman 1999a, b; Theisen &

Brandsegg 2007). This is also the underlying assumption of the second hypothesis of

this thesis, but the results of the analysis contradict these expectations, given the

significance and negative effects of renewable resource scarcity in relation to non-state

conflict35

. Rural population stays interchangeably significant, or close to significant,

with positive effect. Though there are no significant interactions, the aid and refugee

variables almost reach a significant level (0.07), having weak negative effects.

35

Although the bar for qualifying as conflict is equally low in both types of conflict, non-state

conflict is, as previously mentioned, does not suppose governmental involvement and thereby

is assumed to more easily qualify as conflict.

60

5.0. Conclusion

This analysis lends little support to the „renewable resource scarcity‟ perspective. The

logistic time-series analysis offers support to the effects of migration pressure on the

outbreak of conventional conflict, but does not find the same connection when

studying non-state conflict. Here; renewable resource scarcity is found to have

significant negative effects, while this relationship is reversed when it comes to

conventional conflict. Non-state conflict seems to give less support to the neo-

Malthusian perspective, given that the crucial argument of population pressure fails to

be a significant factor when measured as an added pressure by refugees.

The first hypothesis assumes that “Migration pressure is particularly likely to increase

the risk of conflict in the context of increasing renewable resource scarcity”. The

results of the analysis indicate that added population pressure, as operationalized in

this thesis, has significant effects on the outbreak of conflict when investigating

conventional conflict. This variable has no effect on non-state conflict, while the

resource scarcity indicator, however, is significant. The rejection of the first hypothesis

is two-fold. First of all, whenever one of these effects is present, with emphasis on

migration pressure or renewable resource scarcity, all other effects disappear, often

with the exception of the control variables. This means that the independent variables

fail to constitute a situation where these both have significant effects on the dependent

variable. Although being individually important, these effects do not appear

coincidentally and thereby invalidating the hypothesis. Secondly, despite indicating

the significance of renewable resource scarcity, the results indicate a complete reverse

of the hypothesis‟ assumption, as renewable resource scarcity in fact reduces the risk

of conflict.

The second hypothesis describes the types of conflict that are anticipated given its

causes of renewable resource scarcity and migration pressure: “Violent conflict

61

resulting from the interaction between migration pressure and renewable resource

scarcity, is of a low-intensity nature, and does not presuppose government

involvement”. The expectations that this second hypothesis puts forth, explains the use

of the PRIO and Uppsala conflict data and its low requirement of conflict intensity36

.

To fully interpret a denial or confirmation of this second hypothesis is a challenge, as

results from the analysis are mixed, and there are not found significant findings in both

explanatory variables (Potential Cropland and Refugees) to be able to construct the

hypothesised context. An indication of conflict intensity, is the assumption that non-

state conflict, by not demanding government involvement, represents even smaller

conflicts. Following this reasoning, resource scarcity should have positive effects on

this type of conflict, but this is not the case. Renewable resource scarcity has negative

effects implying that such context in fact reduces the chances of non-state conflict.

This betrays the theoretical notion that the greater resource scarcity, the smaller the

conflict. Migration has a significant and positive effect on state-based conflict.

Although this is in line with the overall theory of this thesis, it may reject this

particular hypothesis. One would expect a higher risk of conflict due to migration

pressure regarding non-state conflict, but this is not the case. It is safe to say that this

hypothesis can be rejected. Not only are there no signs of increased risk of non-state

conflict, regarding migration pressure, renewable resource scarcity has either none or

negative effects on this type of conflict.

The third hypothesis predicts that aid has a mediating effect on conflict in the context

of renewable resource scarcity and migration pressure. The supposed reduction in risk

of conflict by aid does not find empirical support. Throughout the different types of

analyses and tests run, the aid variable fails to reach significance. Neither in context,

as supposed by the hypothesis, nor alone does this variable have significant effects.

What this extends to indicate is not easily determined. Empirically the hypothesis is

obviously rejected. However, that this variable remains insignificant may well be due

36

A reduced limit to 25 battle-related deaths from the traditional measure of 1000.

62

to the amounts of missing. If this is the case, a positive effect should be expected (as

the variable in the analysis is inversed), but it is impossible to predict the direction of

effects, precisely because of its insignificance.

In sum, none of the hypotheses, derived from the neo-Malthusian renewable resource

scarcity literature, are empirically supported by the analysis conducted in this thesis.

Methodologically, not only do the different model designs offer different results, but

the analysis is naturally restricted by the data that is used. The time-series analysis

seems to demonstrate effects indicating that the two different kinds of conflict

analysed here are influenced by different factors. The onset of conventional conflict

seems to rely positively on the additional population pressure presented by incoming

refugees. The incidence of non-state conflict on the other hand is rather strongly

influenced by renewable resource scarcity. Remarkably, this entails an effect quite

opposite of what was expected, which reduces chances of conflict by increase in

resource scarcity. The hypotheses suppose interaction mechanisms that are not present

in any of the models, as the variables indicating renewable resource scarcity and

migration pressure never appear significant simultaneously. The lack of clear and

unambiguous findings does not necessarily contradict Homer-Dixon, as he expected

the importance of environmental factors on conflict to increase with time.

[…] As yet, environmental scarcity is not a major factor behind most of these

conflicts, but we can expect it to become a more important influence in coming

decades because of larger populations and higher per capita resource

consumption rates (Homer-Dixon 1999: 13).

This could offer a theoretical explanation for the mixed findings of this analysis, yet

inspire future studies, as the effects of the renewable resource scarcity predicaments

are anticipated to further develop with time and become increasingly important. My

63

findings, however, seem to contradict this statement, as for now, suggesting that

renewable resource scarcity has no impact on conventional conflict, and is associated

with reducing risk of non-state conflict. Further, added population pressure is relevant

only to the outbreak of conventional conflict, while having no effects on non-state

conflict.

The findings of this particular study can arguably have been shaped by both technical

and theoretical factors. The concepts of the resource scarcity and the neo-Malthusian

literature are complex, and thereby perhaps not easily captured in a limited

quantitative study such as this. Contingent arguments and intricate causal chains leave

much to be approximated. This analysis offers a different way of operationalizing

population pressure from the mere conventional measures of the neo-Malthusian

assumption of population growth. The population pressure impact of refugees can be

understood as of more acute importance to an area, relative to natural population

growth. This allows a rather detailed measure of the pressure impact in time, which is

important considering the accurate dating of conflict. Renewable resource scarcity is in

the analysis understood as the relative availability of arable land, obviously excluding

other relevant forms of renewable resources. These are, nevertheless, all necessary

limitations. For better information about the de facto relationships between migration

pressure, resource scarcity, and conflict, a quantitative disaggregated local-level study

is advised. These variables fail to be more area-specific than what is defined by

country boarders. This limits the possibility of connecting population pressure and

local renewable resource scarcity to conflicts in smaller concrete areas and thereby

being able to more reasonably assume that these factors are closely associated.

However, data allowing this is limited, and the manner in which it is undertaken here

should be considered to be a qualified first attempt.

I justified my study by the use of new data, and the alternative causal connections I

investigate by including migration, scarcity and both conventional and non-state

64

internal conflict in my analysis. The interaction of migration and resource scarcity in

potentially causing conflict, have so far not been thoroughly investigated, which leaves

room for the contribution of this Master‟s Thesis. As mentioned, few large-N studies

have found strong empirical support for the relationship between resource scarcity and

internal conflict. The results of this thesis have been able to confirm such a

relationship, although what is remarkable is that the effect is negative, and rather

strongly so. It is safe to say that this represents quite the opposite direction of what

was anticipated. That renewable resource scarcity in fact reduces the risk of conflict

not only rejects a hypothesis, but contradicts the theoretic assumption this was built

on. Despite the Homer-Dixon „clause‟ of a scarcity-violence phenomena in

development, and being increasingly significant, one should not expect such strong

empirical effects suggesting the opposite, but rather expect to find what he himself has

found in his much debated case studies (1999). Migration pressure results offer some

support to the neo-Malthusian perspective, at least concerning conventional conflict,

which traditionally has been the type of conflict under investigation.

Despite an almost unambiguous rejection of all hypotheses, and being aware of the

limitations of the analysis, this thesis may lend some discredit to the most grim

predictions of resource scarcity‟s deterministic violent consequence. Migration,

however, continues to play part in the onset of conventional conflict. These findings

indicate that although environmental degradation, resulting in renewable resource

scarcity, in it self may be no threat, then migration may be. This could support the

„contagion‟ effect in reference to refugees in particular, as discussed, or have

indications of a situation more directly related to environmental issues. Despite not

being concerned with the reasons for flight in this study, and with emphasis on the

great difficulties of being able to map and estimate environmental refugees, these

refugees may, after all, be indirectly relevant to the results of this study. As

environmental degradation, and thereby renewable resource scarcity, causes migration

and not conflict, this could nevertheless ultimately result in internal armed conflict in

the host area, given that migration by itself is found to be a catalyst for conflict.

65

6.0. Bibliography

Barnett, Jon (2001). “Security and Climate Change”. Tyndall Centre Working Paper

No.7. (Downloaded from

http://waterwiki.net/images/4/4e/Security_and_Climate_change.pdf, last accessed

13.04.10)

Black, Richard; Kniveton, Dominic; Skeldon, Ronald; Coppard, Daniel; Murata, Akira

& Schmidt-Verkerk, Kerstin (2008). “Demographics and Climate Change: Future

Trends and their Policy Implications for Migration”. Report prepared for the

Department for International Development by the Sussex Centre for Migration

Research.

Black, Richard & Sessay, Mohamed (1997). “Forced Migration, Environmental

Change and Woodfuel Issues in the Senegal River Valley”. Environmental

Conservation. 24(3): 251–260.

Boserup, Ester (1965). “The Conditions for Agricultural Growth: the Economics of

Agrarian Change under Population Pressure”. Chicago: Aldine.

Boserup, Ester (1981). Population and Technological Change: A Study of Long Term

Trends. Chicago: Chicago University Press.

Brown, Oli (2007). “Climate Change and Forced Migration: Observations, Projections

and Implications”. Human Development Report. (17).

66

Brown, Oli & Crawford, Alec (2008). “Growing Unrest: the Link between Farmed and

Fished Resources and the Risk of Conflict”. Paper. (Downloaded from

http://www.iisd.org/pdf/2008/growing_unrest_resources.pdf, last accessed: 12.04.10).

Buhaug, Halvard & Kristian Skrede Gleditsch (2008.) “Contagion or Confusion? Why

Conflicts Cluster in Space”. International Studies Quarterly 52(2).

Buhaug, Halvard; Nils Petter Gleditsch & Ole Magnus Theisen (2008). “Climate

Change, the Environment and Armed Conflict”. Paper presented at International

Studies Association, San Francisco, CA, 26–29 March (Downloaded from

http://www.allacademic.com/meta/p254278_index.html, last accessed 09.10.09).

Castles, Stephen & Miller, Mark J. (2005). “The Age of Migration: International

Population Movements in the Modern World”. American Foreign Policy Interests.

27(6): 537-542.

Deudney, Daniel (1990). “The Case against Linking Environmental Degradation and

National Security”. Millennium – Journal of International Studies. 19: 461-476.

El-Hinnawi, Essam (1985). Environmental Refugees. Nairobi: UNEP

Eriksson, Mikael & Wallensteen, Peter (2004). “Armed Conflict, 1989-2003”. Journal

of Peace Research. 41(5): 625-636.

67

Forsberg, Erika (2009a). “Neighbours at risk: A Quantitative Study of Civil War

Contagion”. Dissertation. Uppsala University.

Forsberg, Erika (2009b). “Refugees and Intrastate Armed Conflict: A Contagion

Process Approach”. In Forsberg, Erika (2009). Neighbours at risk: A Quantitative

Study of Civil War Contagion. Uppsala University.

Gleditsch, Kristian Skrede (2002) ”Expanded Trade and GDP Data”, Journal of

Conflict Resolution 46(5): 712–724.

Gleditsch, Nils Petter (1998). “Armed Conflict and the Environment: A Critique of the

Literature.” Journal of Peace Research. 35(3): 381-400.

Gleditsch, Nils Petter (2001): ”Resource and Environmental Conflict: The State of the

Art”. In Petzold-Bradley, Eileen; Carius, Alexander & Vincze, Arpad: Responding to

Environmental Conflicts: Implications for Theory and practice. Dordrecht: Kluwer

Academic Publishers. (1-15).

Gleditsch, Nils Petter & Hegre, Håvard (1997). ”Peace and Democracy: Three levels

of Analysis”. Journal of Conflict Resolution. 41(2): 283-310.

Gleditsch, Kristian Skrede and Salehyan, Idean (2006). “Refugees and the Spread of

Civil War.” International Organisation 60: 335-366

68

Gleditsch, Nils Petter & Urdal, Henrik (2002). “Ecoviolence? Links between

Population-Growth, Environmental Scarcity and Violent Conflict in Thomas Homer-

Dixon's Work”. Journal of International Affairs. 56(1): 283–302.

Gleditsch, Nils Petter; Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg &

Håvard Strand (2002). ”Armed Conflict 1946–2001: A New Dataset”. Journal of

Peace Research 39(5): 615–637.

Goldstone, Jack A. (2001). Demography, environment, and security. In: Diehl, Paul F.

& Gleditsch, Nils Petter (Eds.). Environmental Conflict. Oxford: Westiview Press (84-

108).

Greenberg, James B. & Park, Thomas K. (1994). “Political ecology”. Journal of

Political Ecology. 1:1-12

Gurr, Ted Robert (1985). “On the Political Consequences of Scarcity and Economic

Decline”. International Studies Quarterly. 29(1): 51-75.

Hauge, Wenche & Ellingsen, Tanja (1998). “Beyond Environmental Scarcity: Causal

Pathways to Conflict”. Journal of Peace Research 35(3): 299-317.

Hauge, Wenche & Ellingsen, Tanja (2001). Causal pathways to conflict. In: Diehl,

Paul F. & Gleditsch, Nils Petter (Eds.). Environmental Conflict. Oxford: Westiview

Press, 36-57.

69

Hegre, Håvard; Ellingsen, Tanja; Scott Gates & Nils Petter Gleditsch (2001). “Toward

a Democratic Civil Peace? Democracy, Political Change, and Civil War, 1816–1992”.

American Political Science Review 95(1): 33–48.

Hellevik, Ottar (2002). Forskningsmetode i sosiologi og statsvitenskap. Oslo:

Universitetsforlaget.

Homer-Dixon, Thomas F. (1991). “On the Threshold: Environmental Changes as

Causes of Acute Conflict”. International Security. 16(2): 76-116.

Homer-Dixon, Thomas F. (1994). ”Environmental Scarcities and Violent Conflict:

Evidence from Cases”. International Security. 19(1): 5-40

Homer-Dixon, Thomas F. (1995). Strategies for Studying Causation in Complex

Ecological Political Systems. Toronto: Project on Environment, Population and

Security, University of College,

Homer-Dixon, Thomas F. (1999). Environment, Scarcity, and Violence. Princeton, NJ:

Princeton University Press.

Homer-Dixon, Thomas F. & Blitt, Jessica (1998). Ecoviolence: Links among

Environment, Population, and Security. New York: Rowman & Littlefield Publishers

70

Hugo, Graeme (1996). “Environmental Concerns and International Migration”.

International Migration Review. 30(1): 105-131

IPCC, 2001. Third Assessment Report. Climate Change (2001). Geneva:

Intergovernmental Panel on Climate Change & Cambridge: Cambridge University

Press. (Downloaded from http://www.grida.no/publications/other/ipcc_tar/, last

accessed 27.04.10)

Kahl, Colin (2006). States, Scarcity and Civil Strife in the Developing World. New

Jersey: Princeton University Press

Kates, Robert W. (1996). “Population, Technology and the Human Environment: A

Thread through Time”. Daedalus. 125(3).

Lee, Shin-wha (2001). Environment Matters: Conflicts, Refugees and International

Relations. Tokyo: World Human Development Institute (WHDI) Press

Lischer, S. , 2003-08-27 "Collateral Damage: Humanitarian Assistance as a Cause of

Conflict" Paper presented at the annual meeting of the American Political Science

Association, Philadelphia Marriott Hotel, Philadelphia, PA Online <.PDF>. 2009-05-

26 from http://www.allacademic.com/meta/p64512_index.html

Marshall, Monty G. and Keith Jaggers (2002). Polity IV Project: Political Regime

Characteristics and Transitions, 1800-2008. Version 2008. (Downloaded from

http://www.systemicpeace.org/polity/polity4.htm).

71

Meadows, Donella; Randers, Jorgen & Meadows, Dennis (1972). The Limits to

Growth: a Report for the Club of Rome’s Project on the predicament of Mankind.

London: Earth Islands Ltd.

Myers, Norman (1986). “The Environmental Dimension to Security Issues”. The

Environmentalist. 6(4): 251-257

Myers, Norman (1992). “Population/Environment Linkages: Discontinuities ahead?”.

Ambio. 21(1): 116-118

Myers, Norman (1997). ”Environmental Refugees”. Population and Environment.

19(2): 167-182.

NIST/SEMATECH e-Handbook of Statistical Methods, (Downloaded from

http://www.itl.nist.gov/div898/handbook/, last accessed 05.03.10)

Norwegian Refugee Council (NRC) Report (2009). Climate Changed: People

Displaced.

OECD, Organisation for Economic Co-operation and Development. Glossary of

Statistical Terms. (Downloaded from http://stats.oecd.org/glossary/index.htm, last

accessed 01.05.10).

72

Ohlsson, Leif (1999). Environment Scarcity and Conflict: A Study of Malthusian

Concerns.

Peluso, Nancy Lee & Watts, Michael (eds.) (2001). Violent Environments. London:

Cornell University Press

Raknerud, Arvid & Hegre, Håvard (1997). “The Hazard of War. Reassessing the

Evidence of the Democratic Peace”. Journal of Peace Research. 34(4): 385-404.

Raleigh, Clionadh; Jordan, Lisa & Salehyan, Idean (2008). Assessing the impact of

climate change on migration and conflict. Paper prepared for the World Bank Seminar

on Exploring the Social Dimensions of Climate Change. (Downloaded from:

http://siteresources.worldbank.org/EXTSOCIALDEVELOPMENT/Resources/SDCC

WorkingPaper_MigrationandConflict.pdf, last accessed 01.01.10).

Raleigh, Clionadh & Urdal, Henrik (2007). “Climate Change, Environmental

Degradation, and Armed Conflict”. Political Geography. 26: 674-694.

Ree, Joppe & Nillesen, Eleonora (2006). “Aiding Violence or Peace? The Impact of

Foreign Aid on the Risk of Civil Conflict in sub-Saharan Africa”. Journal of

Development Economics 88(2): 301-313.

Reuveny, Rafael (2007). Climate Change-Induced Migration and Conflict. Political

Geography. 26(6): 656–673.

73

Robbins, Paul (2004). Political Ecology: a Critical Introduction. Oxford: Blackwell

Publishing.

Salehyan, Idean (2005). “Human Security and Climate Change”. International

Workshop at Holmen Fjordhotel, Asker June 21-23.

Salehyan, Idean [Distributor] & Gleditsch, Kristian Skrede (2007). "Replication data

for: Refugees and the Spread of Civil War". Data were downloaded from:

http://dvn.iq.harvard.edu/dvn/dv/salehyan/faces/study/StudyPage.xhtml?studyId=658,

last accessed 10.11.09

Schwartz, Daniel M.; Delogiannis, Tom & Homer-Dixon, Thomas (2001). The

environment and violent conflict. In: Diehl, Paul F. & Gleditsch, Nils Petter (Eds.).

Environmental Conflict. Oxford: Westview Press, 273-294.

Simon, Julian (1981). The Ultimate Resource. Princeton: Princeton University Press

Simon, Julian (1996). The Ultimate Resource 2. New Jersey: Princeton University

Press.

Skog, Ole-Jørgen (2007). Å forklare sosiale fenomener: En regresjonsbasert

tilnærming. Oslo: Gyldendal Akademisk

74

de Soysa, Indra (2002). “Ecoviolence: Shrinking Pie or Honey Pot?”. Global

Environmental Politics 2(4): 1–34.

de Soysa, Indra (2005). “Filthy Rich, Not Dirt Poor! How Nature Nurtures Civil

Violence” In Peter Dauvergne (Ed.) Handbook of Environmental Politics. London:

Edward Elgar (149–169).

StatSoft. Electronic Statistical Textbook. (Downloaded from

http://www.statsoft.com/textbook/time-series-analysis/, last accessed 07.05.10).

Strand, Håvard (2006). 'Onset of Armed Conflict: A New List for the period 1946–

2004, with Applications', under review at Conflict Management and Peace Science.

(http://www.prio.no/upload/983/Onset.pdf, last accessed 01.03.10)

Suliman, Mohamed (1999a). ”Conflict Resolution among the Borana and the Fur:

Similar Features, Different Outcomes”. In Mohamed Suliman (ed.), Ecology, Politics

and Violent Conflict. London: Zed, 286–289.

Suliman, Mohamed (1999b). “The Rationality and Irrationality of Violence in Sub-

Saharan Africa”. In: Mohamed Suliman (ed.) Ecology, Politics and Violent Conflict.

London: Zed, 27–43.

Suhrke, Astri (1991). “Pressure Points: Environmental Degradation, Migration and

Conflict”. Paper for the workshop on “Environmental Change, Population

Displacement, and Acute Conflict held at the Institute for Research on Public Policy,

75

in Ottawa June 1991. (Downloaded from

http://www.cmi.no/publications/1993%5Cpressure_points.pdf, last accessed 11.04.10).

Suhrke, Astri (1993). “Pressure points: environmental degradation, migration and

conflict”. In Workshop on Environmental Change, Population Displacement, and

Acute Conflict. Institute for Research on Public Policy, Ottawa.

Suhrke, Astri (1997). “Environmental Degradation, Migration, and the Potential for

Violent Conflict”. In Gleditsch (1997). Conflict and the Environment. London: Kluwer

Academic Publishers.

Tarp, Finn (ed.) & Hjertholm, Peter (2000). Foreign Aid and Developments: Lessons

Learnt and Directions for the Future. New York: Rutledge

Theisen, Ole Magnus (2006). ”Renewable Resource Scarcity and Internal Armed

Conflict: A Quantitative Reassessment”. Thesis for the MA degree in Political Science.

Department of Sociology and Political Science. Trondheim: NTNU

Theisen, Ole Magnus (2008). ”Blood and Soil? Resource Scarcity and Internal Armed

Conflict Revisited”. Journal of Peace Research. 45(6): 801-818.

Theisen, Ole Magnus & Brandsegg, Kristian Bjarnøe (2007). “The Environment and

Non-State Conflict”. Presented at Statsvitenskapelig fagkonferanse, Trondheim 3-5

January.

76

Tir, Jaroslav & Diehl, Paul F. (2001). Demographic pressure and interstate conflict. In:

Diehl, Paul F. & Gleditsch, Nils Petter (Eds.). Environmental Conflict. Oxford:

Westiview Press, 58-83

UCDP/PRIO Armed Conflict Dataset Codebook Version 4 (2009). (Downloaded from

http://www.pcr.uu.se/research/UCDP/data_and_publications/datasets.htm, last

accessed 04.03.10)

UCDP Non-State Conflict Dataset Codebook Version 2.1 (2009). (Downloaded from

http://www.pcr.uu.se/research/UCDP/data_and_publications/datasets.htm, last

accessed 04.03.10)

UN, 2008. Urban and Rural Population by Age and Sex, 1950-2005. Version 1. New

York: United Nations Population Division. CD-ROM May 2008.

UNHCR (1951). Convention and Protocol Relating to the Status of Refugees. Geneva:

United Nations High Commissioner for Refugees.

UNHCR (1996). Environmental Guidelines. Geneva: United Nations High

Commissioner for Refugees.

United States Committee for Refugees and Immigrants (USCRI) (Annual Series).

“World Refugee Survey Forcibly Displaced Populations, 1964-2008”. Compiled by

Monty G. Marshall, Center for Systemic Peace (Downloaded from

www.systemicpeace.org, last accessed 20.01.10)

77

Unruh, Krol and Kliot (2004). Environmental Change and its Implications for

Population Migration. London: Kluwer Academic Publishers.

Urdal, Henrik (2005). “People vs. Malthus: Population Pressure, Environmental

Degradation, and Armed Conflict Revisited.” Journal of Peace Research 42(4): 417-

434.

Weiner, Myron (1993). International Migration and Security. Boulder, Colo.

Westview Press.

Weiner, Myron (1995). The Global Migration Crisis: Challenge to States and to

Human Rights. New York: Harper Collins.

Weiner, Myron (1996). “Bad Neighbours, Bad Neighbourhoods: An Inquiry into the

Causes of Refugee Flows”. International Security. 21(1): 5-42.

Weiner, Myron & Russell (2001). Demography and National Security. New York:

Berghahn Books.

Weiner, Myron & Teitelbaum, Michael T. (2001). Political Demography,

Demographic Engineering. New York: Berghahn Books.

78

Welch, Susan & Miewald, Robert (1983). Scarce Natural Resources: the Challenge to

Public Policymaking. Beverly Hills: Sage Publications.

Wood, William (1994). “Forced Migration: Local Conflicts and International

Dilemmas”. Annals of the Association of American Geographers. 84(4): 607-634.

World Bank (2008). World Development Indicators 2008 [CD-ROM]. Washington,

DC: World Bank.

Zolberg, Aristide R. & Benda, Peter (2001). Global Migrants, Global Refugees:

Problems and Solutions. New York: Berghahn Books.

79

7.0. Appendix

7.1. Descriptive Statistics over the Dependent Variables

Variables Obs Mean Std. Dev. Min Max

onset 7095 0.028189 0.165524 0 1

NonStCon 1013 0.057256 0.232445 0 1

7.2. Variable Correlation Matrix37

ConvC Ref AidP PotC RurPo TotPo BrevP IMR Reg RegSq RegM

ConvC 1

Ref 0.0522 1

AidP -0.01 -0.008 1

PotC -0.011 -0.074 0.0197 1

RurPo 0.0683 -0.044 0.0387 -0.201 1

TotPo 0.0415 0.3704 -0.36 0.1263 -0.045 1

BrevP -0.016 0.1146 0.0265 0.0378 0.2011 0.2598 1

IMR 0.0796 -0.066 0.0438 -0.296 0.724 -0.122 0.2024 1

Reg -0.036 0.0396 -0.06 0.069 -0.36 0.1436 -0.073 -0.533 1

RegSq -0.085 -0.077 -0.116 0.0319 -0.369 -0.005 -0.27 -0.401 0.3123 1

RegM -0.003 0.0205 0.1038 0.0529 -0.001 -0.028 0.0845 0.0016 -0.008 -0.165 1

Conventional Conflict, 1964-2007 (CSP data)

37

Multicollinearity is difficult to reject. A rule of thumb is that correlations between two

variables of 0.8 or higher could indicate a problem of multicollinearity. A sign of this problem

is that model results are very unstable, depending on the inclusion or exclusion of the highly

correlated variables. The correlation matrix in this analysis, however, indicates no significant

correlation. The variables that are most highly correlated are the development indicators; IMR

and GDP, but these are never analysed simultaneously.

80

ConvC Ref AidP PotC RurPo TotPo BrevP IMR Reg RegSq RegM

ConvC 1

Ref 0.0599 1

AidP -0.015 0.0137 1

PotC -0.013 -0.003 0.0569 1

RurPo 0.055 0.1448 -0.006 -0.238 1

TotPo 0.0502 0.2817 -0.369 0.0863 0.004 1

BrevP -0.015 0.1837 0.0152 0.0386 0.1715 0.2731 1

IMR 0.0668 0.0539 -0.026 -0.306 0.7396 -0.052 0.1674 1

Reg -0.034 -0.101 -0.052 0.0746 -0.354 0.122 -0.056 -0.533 1

RegSq -0.073 -0.134 -0.109 0.0149 -0.289 0.0777 -0.218 -0.314 0.2616 1

RegM -0.026 -0.099 0.2214 0.0926 -0.124 -0.288 -0.019 -0.106 0.0013 -0.319 1

Conventional Conflict, 195-2001 (Salehyan & Gleditsch data)

ConvC Ref AidP PotC RurPo TotPo BrevP IMR Reg RegSq RegM

ConvC 1

Ref 0.1263 1

AidP -0.053 -0.075 1

PotC -0.019 -0.011 -0.134 1

RurPo 0.2028 -0.004 0.2153 -0.03 1

TotPo 0.236 0.3907 -0.409 0.154 0.0045 1

BrevP 0.309 0.1931 -0.028 0.0617 0.2934 0.3175 1

IMR 0.2569 0.1883 0.2306 -0.198 0.626 -0.054 0.2976 1

Reg -0.051 -0.081 -0.038 -0.05 -0.23 0.0545 -0.111 -0.322 1

RegSq -0.198 -0.196 -0.309 -0.002 -0.424 0.0145 -0.281 -0.625 0.5086 1

RegM -0.029 0.0423 0.1971 0.0887 -0.012 -0.068 -0.021 -0.062 -0.058 -0.179 1

Non-State Conflict, 2002-2007 (CSP data)

81

7.3. Logistic Regression Time-Series Results in Coefficients

Onset Coef. Std. Err. z P>z [95% Conf. Interval]

Refugees .0505199 .0154538 3.27 0.001 .020231 .0808088

IMR .0055486 .0015862 3.50 0.000 .0024397 .0086575

TotPopLn .1941538 .0518421 3.75 0.000 .0925451 .2957624

Regime .0039348 .013548 0.29 0.771 -.0226188 .0304884

RegimeSq -.0153245 .0027685 -5.54 0.000 -.0207507 -.0098983

RegimeMiss -2.283418 1.026361 -2.22 0.026 -4.295048 -.271787

BrevPeace 1.242742 .2998043 4.15 0.000 .6551359 1.830347

_cons -5.231239 .5352178 -9.77 0.000 -6.280246 -4.182231

N = 5921, LR chi2(7) =139.40, Prob > chi2 = 0.0000, Log likelihood = -718.94684, Pseudo R2 = 0.0884

Salehyan & Gleditsch data, basic model (coefficients). Conventional conflict.

Onset Coef. Std. Err. z P>z [95%Conf. Interval]

Refugees .0430273 .0173813 2.48 0.013 .0089606 .0770941

Potential C. -.0192887 .0632243 -0.31 0.760 -.143206 .1046286

Aidpercap .000323 .0028818 0.11 0.911 -.0053251 .0059711

RurPop -.0012863 .0055138 -0.23 0.816 -.0120931 .0095205

IMR .0073313 .0024982 2.93 0.003 .0024349 .0122278

TotPopLn .2452529 .0682169 3.60 0.000 .1115503 .3789555

Regime .003687 .0150751 0.24 0.807 -.0258596 .0332335

RegimeSq -.0146209 .003015 -4.85 0.000 -.0205302 -.0087115

RegimeMiss -1.280297 1.05535 -1.21 0.225 -3.348745 .7881513

BrevPeace 1.19367 .3252719 3.67 0.000 .5561491 1.831192

_cons -5.642531 .7305966 -7.72 0.000 -7.074474 -4.210588

N = 4606, LR chi2(10) = 115.41, Prob > chi2 = 0.0000, Log likelihood = -596.69987, Pseudo R2 = 0.0882

Salehyan & Gleditsch data, all variables included (coefficients). Conventional conflict.

7.4. “Do-File” (STATA 9 )

[Note: all commands yet not all run tests are shown below. This is particularly so for instance

in relation to the alternative years checked in the cross-sectional analysis.]

*GETTING STARTED

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamain.dta", clear

rename countryname country

82

rename imr2 imr

*UPDATING POTENTIAL CROPLAND VARIABLE

gen temp = .

replace temp = u_prodarea if year == 2000

sort country year

by country: egen temp2 = mean(u_prodarea)

replace u_popdprod = 1000*totpop/temp2 if year > 2000

*NEW VARIABLES

*Make refugees (CSP) per capita

gen cspref_pc=csp_host/totpop

*Make refugees (Salehyan & Gleditsch) per capita

gen s_logtotrefh_pc= s_logtotrefh/totpop

*Make logtransformed CSP refugee data

gen csp_temp=csp_host

replace csp_temp=0.001 if csp_temp==0

gen csp_loghost=log(csp_temp)

* Make logtransformed potential cropland variable

gen u_popdprodln= ln(u_popdprod)

save "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta", replace

********************************** CORRELATIONS **************************

************************ ALL VARIABLES CORRELATION MATRIX ****************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(1) VARIABLE CORRELATION CSP REF. CONVENTIONAL CONFLICT

corr onset csp_loghost aidpercap u_popdprodln wup_prosrur totpopln

brevpeace imr polity politysq polmiss

corr onset csp_loghost aidpercap u_popdprodln wup_prosrur totpopln

brevpeace ksg_gdppc polity politysq polmiss

*(2) VARIABLE CORRELATION SALEHYAN & GLEDITSCH REF. CONVENTIONAL CONFLICT

corr onset s_logtotrefh aidpercap u_popdprodln wup_prosrur totpopln

brevpeace imr polity politysq polmiss

corr onset s_logtotrefh aidpercap u_popdprodln wup_prosrur totpopln

brevpeace ksg_gdppc polity politysq polmiss

*(3) VARIABLE CORRELATION CSP REF. NON-STATE CONFLICT

corr nonscinc csp_loghost aidpercap u_popdprodln wup_prosrur totpopln

brevpeace imr polity politysq polmiss

corr nonscinc csp_loghost aidpercap u_popdprodln wup_prosrur totpopln

brevpeace ksg_gdppc polity politysq polmiss

****** MODEL CORRELATION (correlation between coefficients) ******

*(1) CORRELATION MODEL REFUGEE DATA (CSP)

83

logistic onset csp_loghost totpopln brevpeace imr polity politysq polmiss

if !(onset==0 & incidence==1)

logistic onset csp_loghost totpopln brevpeace ksg_gdppc polity politysq

polmiss if !(onset==0 & incidence==1)

vce, corr

*(2) CORRELATION MODEL REFUGEE DATA (Salehyan & Gleditsch)

logit onset s_logtotrefh totpopln brevpeace imr polity politysq polmiss if

!(onset==0 & incidence==1), or

logit s_logtotrefh csp_host totpopln brevpeace ksg_gdppc polity politysq

polmiss if !(onset==0 & incidence==1), or

vce, corr

*(3) CORRELATION MODEL REFUGEE DATA, NON-STATE (CSP)

logit nonscinc csp_loghost totpopln brevpeace imr polity politysq polmiss,

or

logit nonscinc csp_loghost totpopln brevpeace ksg_gdppc polity politysq

polmiss, or

vce, corr

***********************************************************************

***********************************************************************

***********************************************************************

************************ BASIC MODELS *********************************

*(1) LOGISTIC BASIC MODEL 1.0. (conventional conflict, CSP ref data)

logistic onset csp_loghost imr totpopln polity politysq polmiss brevpeace

if !(onset==0 & incidence==1)

***

*(1) LOGISTIC ALTERNATIVE MODEL

*Contecstual variables:

logistic onset csp_loghost u_popdprodln aidpercap wup_prosrur imr totpopln

polity politysq polmiss brevpeace if !(onset==0 & incidence==1)

*(2) LOGISTIC BASIC MODEL 2.0. (conventional conflict, Salehyan and

Gleditsch ref data)

logistic onset s_logtotrefh imr totpopln polity politysq polmiss brevpeace

if !(onset==0 & incidence==1)

***

*(2) LOGISTIC ALTERNATIVE MODEL

*Contecstual variables:

logistic onset s_logtotrefh u_popdprodln aidpercap wup_prosrur imr totpopln

polity politysq polmiss brevpeace if !(onset==0 & incidence==1)

84

*(3) LOGISTICBASIC MODEL 3.0. (non-state conflict, CSP ref data)

logistic nonscinc csp_loghost imr totpopln polity politysq polmiss

brevpeace

***

*(3) LOGISTIC ALTERNATIVE MODEL

*Contecstual variables:

logistic nonscinc csp_loghost u_popdprodln aidpercap wup_prosrur imr

totpopln polity politysq polmiss brevpeace

**************************************************************************

**************************************************************************

******************** MODELS WITH VARIABLES OF INTERACTION ****************

*Run the following when appropriate:

*Goodness-of-fit test (-> Pearsons Chi square not appropriate)

lfit

*Goodness-of-fit test

lfit, group(10) table

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(1) MODEL 1.1. (refugees & share of rural population)

*generate interaction

sum csp_loghost

gen csp_loghost_c=csp_loghost-1.655

sum wup_prosrur

gen wup_prosrur_c=wup_prosrur-55.223

gen csp_loghostprosrur=csp_loghost_c*wup_prosrur_c

*model with interaction

logistic onset csp_loghost_c csp_loghostprosrur wup_prosrur_c imr totpopln

polity politysq polmiss brevpeace if !(onset==0 & incidence==1)

logistic onset csp_loghost_c csp_loghostprosrur u_popdprodln aidpercap

wup_prosrur_c imr totpopln polity politysq polmiss brevpeace if !(onset==0

& incidence==1)

***

logistic onset csp_loghost_c csp_loghostprosrur wup_prosrur_c ksg_gdppc

totpopln polity politysq polmiss brevpeace if !(onset==0 & incidence==1)

logistic onset csp_loghost_c csp_loghostprosrur u_popdprodln aidpercap

wup_prosrur_c ksg_gdppc totpopln polity politysq polmiss brevpeace if

!(onset==0 & incidence==1)

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

85

*(1) INTERACTION MODEL 1.2. (refugees & potential cropland)

*generate interaction

sum csp_loghost

gen csp_loghost_c=csp_loghost-1.655

sum u_popdprodln

gen u_popdprodln_c=u_popdprodln-4.177

gen csp_loghostpopdprod=csp_loghost_c*u_popdprodln_c

*model with interaction

logistic onset csp_loghost_c u_popdprodln_c csp_loghostpopdprod totpopln

polity politysq polmiss imr brevpeace if !(onset==0 & incidence==1)

logistic onset csp_loghost_c u_popdprodln_c csp_loghostpopdprod aidpercap

wup_prosrur totpopln polity politysq polmiss imr brevpeace if !(onset==0 &

incidence==1)

***

logistic onset csp_loghost_c u_popdprodln_c csp_loghostpopdprod totpopln

polity politysq polmiss ksg_gdppc brevpeace if !(onset==0 & incidence==1)

logistic onset csp_loghost_c u_popdprodln_c csp_loghostpopdprod aidpercap

wup_prosrur totpopln polity politysq polmiss ksg_gdppc brevpeace if

!(onset==0 & incidence==1)

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(1) INTERACTION MODEL 1.3. (refugees & aid)

*generate interaction

sum aidpercap

gen aidpercap_c=aidpercap-27.697

sum csp_loghost

gen csp_loghost_c=csp_loghost-1.655

gen csp_loghostaidpercap=csp_loghost_c*(aidpercap_c*-1)

*model with interaction

logistic onset csp_loghost_c aidpercap_c csp_loghostaidpercap totpopln

polity politysq polmiss imr brevpeace if !(onset==0 & incidence==1)

logistic onset csp_loghost_c aidpercap_c csp_loghostaidpercap wup_prosrur

u_popdprodln totpopln polity politysq polmiss imr brevpeace if !(onset==0 &

incidence==1)

***

logistic onset csp_loghost_c aidpercap_c csp_loghostaidpercap totpopln

polity politysq polmiss ksg_gdppc brevpeace if !(onset==0 & incidence==1)

logistic onset csp_loghost_c aidpercap_c csp_loghostaidpercap wup_prosrur

u_popdprodln totpopln polity politysq polmiss ksg_gdppc brevpeace if

!(onset==0 & incidence==1)

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

86

*(1) INTERACTION MODEL 1.4. (potential cropland & share of rural

population)

*generate interaction

sum u_popdprodln

gen u_popdprodln_c=u_popdprodln-4.177

sum wup_prosrur

gen wup_prosrur_c=wup_prosrur-55.223

gen prosrurpopdprod=(u_popdprodln_c*-1)*wup_prosrur_c

*model with interaction

logistic onset csp_loghost prosrurpopdprod u_popdprodln_c wup_prosrur_c imr

totpopln polity politysq polmiss brevpeace if !(onset==0 & incidence==1)

logistic onset csp_loghost prosrurpopdprod u_popdprodln_c aidpercap

wup_prosrur_c imr totpopln polity politysq polmiss brevpeace if !(onset==0

& incidence==1)

***

logistic onset csp_loghost prosrurpopdprod u_popdprodln_c wup_prosrur_c

ksg_gdppc totpopln polity politysq polmiss brevpeace if !(onset==0 &

incidence==1)

logistic onset csp_loghost prosrurpopdprod u_popdprodln_c aidpercap

wup_prosrur_c ksg_gdppc totpopln polity politysq polmiss brevpeace if

!(onset==0 & incidence==1)

**************************************************************************

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(2) INTERACTION MODEL 2.1. (refugees & share of rural population)

*generate interaction

sum s_logtotrefh

gen s_logtotrefh_c=s_logtotrefh-2.708

sum wup_prosrur

gen wup_prosrur_c=wup_prosrur-55.223

gen s_logtotrefhprosrur=s_logtotrefh_c*wup_prosrur_c

*model with interaction

logistic onset s_logtotrefh_c s_logtotrefhprosrur wup_prosrur_c imr

totpopln polity politysq polmiss brevpeace if !(onset==0 & incidence==1)

logistic onset s_logtotrefh_c s_logtotrefhprosrur u_popdprodln aidpercap

wup_prosrur_c imr totpopln polity politysq polmiss brevpeace if !(onset==0

& incidence==1)

***

logistic onset s_logtotrefh_c s_logtotrefhprosrur wup_prosrur_c ksg_gdppc

totpopln polity politysq polmiss brevpeace if !(onset==0 & incidence==1)

logistic onset s_logtotrefh_c s_logtotrefhprosrur u_popdprodln aidpercap

wup_prosrur_c ksg_gdppc totpopln polity politysq polmiss brevpeace if

!(onset==0 & incidence==1)

*************************************************************************

87

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(2) INTERACTION MODEL 2.2. (refugees & potential cropland)

*generate interaction

sum s_logtotrefh

gen s_logtotrefh_c=s_logtotrefh-2.708

sum u_popdprodln

gen u_popdprodln_c=u_popdprodln-4.177

gen s_logtotrefhpopdprod=s_logtotrefh_c*u_popdprodln_c

*model with interaction

logistic onset s_logtotrefh_c u_popdprodln_c s_logtotrefhpopdprod totpopln

polity politysq polmiss imr brevpeace if !(onset==0 & incidence==1)

logistic onset s_logtotrefh_c u_popdprodln_c s_logtotrefhpopdprod aidpercap

wup_prosrur totpopln polity politysq polmiss imr brevpeace if !(onset==0 &

incidence==1)

***

logistic onset s_logtotrefh_c u_popdprodln_c s_logtotrefhpopdprod totpopln

polity politysq polmiss ksg_gdppc brevpeace if !(onset==0 & incidence==1)

logistic onset s_logtotrefh_c u_popdprodln_c s_logtotrefhpopdprod aidpercap

wup_prosrur totpopln polity politysq polmiss ksg_gdppc brevpeace if

!(onset==0 & incidence==1)

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(2) INTERACTION MODEL 2.3. (refugees & aid)

*generate interaction

sum aidpercap

gen aidpercap_c=aidpercap-27.697

sum s_logtotrefh

gen s_logtotrefh_c=s_logtotrefh-2.708

gen s_logtotrefhaidpercap=s_logtotrefh_c*(aidpercap_c*-1)

*model with interaction

logistic onset s_logtotrefhaidpercap s_logtotrefh_c aidpercap_c aidpercap

totpopln polity politysq polmiss imr brevpeace if !(onset==0 &

incidence==1)

logistic onset s_logtotrefhaidpercap s_logtotrefh_c aidpercap_c wup_prosrur

u_popdprodln totpopln polity politysq polmiss imr brevpeace if !(onset==0 &

incidence==1)

***

logistic onset s_logtotrefhaidpercap s_logtotrefh_c aidpercap_c totpopln

polity politysq polmiss ksg_gdppc brevpeace if !(onset==0 & incidence==1)

88

logistic onset s_logtotrefhaidpercap s_logtotrefh_c aidpercap_c wup_prosrur

u_popdprodln totpopln polity politysq polmiss ksg_gdppc brevpeace if

!(onset==0 & incidence==1)

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(2) INTERACTION MODEL 2.4. (potential cropland & share of rural

population)

*generate interaction

sum u_popdprodln

gen u_popdprodln_c=u_popdprodln-4.177

sum wup_prosrur

gen wup_prosrur_c=wup_prosrur-55.223

gen prosrurpopdprod=(u_popdprodln_c*-1)*wup_prosrur_c

*model with interaction

logistic onset s_logtotrefh prosrurpopdprod u_popdprodln_c wup_prosrur_c

imr totpopln polity politysq polmiss brevpeace if !(onset==0 &

incidence==1)

logistic onset s_logtotrefh prosrurpopdprod u_popdprodln_c aidpercap

wup_prosrur_c imr totpopln polity politysq polmiss brevpeace if !(onset==0

& incidence==1)

***

logistic onset s_logtotrefh prosrurpopdprod u_popdprodln_c wup_prosrur_c

ksg_gdppc totpopln polity politysq polmiss brevpeace if !(onset==0 &

incidence==1)

logistic onset s_logtotrefh prosrurpopdprod u_popdprodln_c aidpercap

wup_prosrur_c ksg_gdppc totpopln polity politysq polmiss brevpeace if

!(onset==0 & incidence==1)

**************************************************************************

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(3) INTERACTION MODEL 3.1. (refugees & share of rural population)

*generate interaction

sum csp_loghost

gen csp_loghost_c=csp_loghost-1.655

sum wup_prosrur

gen wup_prosrur_c=wup_prosrur-55.223

gen csp_loghostprosrur=csp_loghost_c*wup_prosrur_c

*model with interaction

logistic nonscinc csp_loghost_c csp_loghostprosrur wup_prosrur_c imr

totpopln polity politysq polmiss brevpeace

logistic nonscinc csp_loghost_c csp_loghostprosrur u_popdprodln aidpercap

wup_prosrur_c imr totpopln polity politysq polmiss brevpeace

***

89

logistic nonscinc csp_loghost_c csp_loghostprosrur wup_prosrur_c ksg_gdppc

totpopln polity politysq polmiss brevpeace

logistic nonscinc csp_loghost_c csp_loghostprosrur u_popdprodln aidpercap

wup_prosrur_c ksg_gdppc totpopln polity politysq polmiss brevpeace

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(3) INTERACTION MODEL 3.2. (refugees & potential cropland)

*generate interaction

sum csp_loghost

gen csp_loghost_c=csp_loghost-1.655

sum u_popdprodln

gen u_popdprodln_c=u_popdprodln-4.177

gen csp_loghostpopdens=csp_loghost_c*u_popdprodln_c

*model with interaction

logistic nonscinc csp_loghost_c u_popdprodln_c csp_loghostpopdens totpopln

polity politysq polmiss imr brevpeace

logistic nonscinc csp_loghost_c u_popdprodln_c csp_loghostpopdens aidpercap

wup_prosrur totpopln polity politysq polmiss imr brevpeace

***

logistic nonscinc csp_loghost_c u_popdprodln_c csp_loghostpopdens totpopln

polity politysq polmiss ksg_gdppc brevpeace

logistic nonscinc csp_loghost_c u_popdprodln_c csp_loghostpopdens aidpercap

wup_prosrur totpopln polity politysq polmiss ksg_gdppc brevpeace

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(3) INTERACTION MODEL 3.3. (refugees & aid)

*generate interaction

sum aidpercap

gen aidpercap_c=aidpercap-27.697

sum csp_loghost

gen csp_loghost_c=csp_loghost-1.655

gen csp_loghostaidpercap=csp_loghost_c*(aidpercap_c*-1)

*model with interaction

logistic nonscinc csp_loghost_c aidpercap_c csp_loghostaidpercap totpopln

polity politysq polmiss imr brevpeace

logistic nonscinc csp_loghost_c aidpercap_c csp_loghostaidpercap

wup_prosrur u_popdprodln totpopln polity politysq polmiss imr brevpeace

***

logistic nonscinc csp_loghost_c aidpercap_c csp_loghostaidpercap totpopln

polity politysq polmiss ksg_gdppc brevpeace

90

logistic nonscinc csp_loghost_c aidpercap_c csp_loghostaidpercap

wup_prosrur u_popdprodln totpopln polity politysq polmiss ksg_gdppc

brevpeace

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(3) INTERACTION MODEL 3.4. (potential cropland & share of rural

population)

*generate interaction

sum u_popdprodln

gen u_popdprodln_c=u_popdprodln-4.177

sum wup_prosrur

gen wup_prosrur_c=wup_prosrur-55.223

gen prosrurpopdprod=(u_popdprodln_c*-1)*wup_prosrur_c

*model with interaction

logistic nonscinc csp_loghost prosrurpopdprod u_popdprodln_c wup_prosrur_c

imr totpopln polity politysq polmiss brevpeace

logistic nonscinc csp_loghost prosrurpopdprod u_popdprodln_c aidpercap

wup_prosrur_c imr totpopln polity politysq polmiss brevpeace

***

logistic nonscinc csp_loghost prosrurpopdprod u_popdprodln_c wup_prosrur_c

ksg_gdppc totpopln polity politysq polmiss brevpeace

logistic nonscinc csp_loghost prosrurpopdprod u_popdprodln_c aidpercap

wup_prosrur_c ksg_gdppc totpopln polity politysq polmiss brevpeace

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(3) INTERACTION MODEL 3.5. (potential cropland & aid)

*generate interaction

sum u_popdprodln

gen u_popdprodln_c=u_popdprodln-4.177

sum aidpercap

gen aidpercap_c=aidpercap-27.697

gen popdprodaid=(u_popdprodln_c*-1)*(aidpercap_c*-1)

logistic nonscinc csp_loghost popdprodaid u_popdprodln_c wup_prosrur

aidpercap_c imr totpopln polity politysq polmiss brevpeace

**************************************************************************

**************************************************************************

********************** CROSS-SECTIONAL ANALYSIS **************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(1) CROSS-SECTIONAL BASIC MODEL 1.0. (conventional conflict, CSP ref data)

91

logistic onset csp_loghost imr totpopln polity politysq polmiss brevpeace

if !(onset==0 & incidence==1) & year==2007

*(1) CROSS-SECTIONAL ALTERNATIVE MODEL

*Contecstual variables (minus aid):

logistic onset csp_loghost u_popdprodln wup_prosrur imr totpopln polity

politysq polmiss brevpeace if !(onset==0 & incidence==1) & year==2007

*(2) CROSS-SECTIONAL BASIC MODEL 2.0. (conventional conflict, Salehyan and

Gleditsch ref data)

logistic onset s_logtotrefh imr totpopln polity politysq polmiss brevpeace

if !(onset==0 & incidence==1) & year==2001

*(2) CROSS-SECTIONAL ALTERNATIVE MODEL

*Contecstual variables (minus aid):

logistic onset s_logtotrefh u_popdprodln wup_prosrur imr totpopln polity

politysq polmiss brevpeace if !(onset==0 & incidence==1) & year==2001

*(3) CROSS-SECTIONAL BASIC MODEL 3.0. (non-state conflict, CSP ref data)

logistic nonscinc csp_loghost imr totpopln polity politysq polmiss

brevpeace if year==2007

*(3) CROSS-SECTIONAL ALTERNATIVE MODEL

*Contecstual variables (minus aid):

logistic nonscinc csp_loghost u_popdprodln wup_prosrur imr totpopln polity

politysq polmiss brevpeace if year==2007

**************************************************************************

********************** CROSS-SECTIONAL INTERACTION **********************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(1) CROSS-SECTIONAL INTERACTION CSP CONVENTIONAL MODEL 1.1. (refugees &

share of rural population)

*generate interaction

sum csp_loghost

gen csp_loghost_c=csp_loghost-1.655

sum wup_prosrur

gen wup_prosrur_c=wup_prosrur-55.223

gen csp_loghostprosrur=csp_loghost_c*wup_prosrur_c

*model with interaction

logistic onset csp_loghost_c csp_loghostprosrur wup_prosrur_c imr totpopln

polity politysq polmiss brevpeace if !(onset==0 & incidence==1) &

year==2007

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

92

*(1) CROSS-SECTIONAL INTERACTION CSP CONVENTIONAL MODEL 1.2. (refugees &

potential cropland)

*generate interaction

sum csp_loghost

gen csp_loghost_c=csp_loghost-1.655

sum u_popdprodln

gen u_popdprodln_c=u_popdprodln-4.177

gen csp_loghostpopdprod=csp_loghost_c*u_popdprodln_c

*model with interaction

logistic onset csp_loghost_c u_popdprodln_c csp_loghostpopdprod totpopln

polity politysq polmiss imr brevpeace if !(onset==0 & incidence==1) &

year==2007

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(1) CROSS-SECTIONAL INTERACTION CSP CONVENTIONAL MODEL 1.3. (refugees &

aid)

*generate interaction

sum aidpercap

gen aidpercap_c=aidpercap-27.697

sum csp_loghost

gen csp_loghost_c=csp_loghost-1.655

gen csp_loghostaidpercap=csp_loghost_c*(aidpercap_c*-1)

*model with interaction

logistic onset csp_loghost_c aidpercap_c csp_loghostaidpercap totpopln

polity politysq polmiss imr brevpeace if !(onset==0 & incidence==1) &

year==2007

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(1) CROSS-SECTIONAL INTERACTION CSP CONVENTIONAL MODEL 1.4. (potential

cropland & share of rural population)

*generate interaction

sum u_popdprodln

gen u_popdprodln_c=u_popdprodln-4.177

sum wup_prosrur

gen wup_prosrur_c=wup_prosrur-55.223

gen prosrurpopdprod=(u_popdprodln_c*-1)*wup_prosrur_c

*model with interaction

logistic onset csp_loghost prosrurpopdprod u_popdprodln_c wup_prosrur_c imr

totpopln polity politysq polmiss brevpeace if !(onset==0 & incidence==1) &

year==2003

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

93

*(2) CROSS-SECTIONAL INTERACTION S&G CONVENTIONAL MODEL 2.1. (refugees &

share of rural population)

*generate interaction

sum s_logtotrefh

gen s_logtotrefh_c=s_logtotrefh-2.708

sum wup_prosrur

gen wup_prosrur_c=wup_prosrur-55.223

gen s_logtotrefhprosrur=s_logtotrefh_c*wup_prosrur_c

*model with interaction

logistic onset s_logtotrefh_c s_logtotrefhprosrur wup_prosrur_c imr

totpopln polity politysq polmiss brevpeace if !(onset==0 & incidence==1) &

year==2001

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(2) CROSS-SECTIONAL INTERACTION S&G CONVENTIONAL MODEL 2.2. (refugees &

potential cropland)

*generate interaction

sum s_logtotrefh

gen s_logtotrefh_c=s_logtotrefh-2.708

sum u_popdprodln

gen u_popdprodln_c=u_popdprodln-4.177

gen s_logtotrefhpopdprod=s_logtotrefh_c*u_popdprodln_c

*model with interaction

logistic onset s_logtotrefh_c u_popdprodln_c s_logtotrefhpopdprod totpopln

polity politysq polmiss imr brevpeace if !(onset==0 & incidence==1) &

year==2001

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(2) CROSS-SECTIONAL INTERACTION S&G CONVENTIONAL MODEL 2.3. (refugees &

aid)

*generate interaction

sum aidpercap

gen aidpercap_c=aidpercap-27.697

sum s_logtotrefh

gen s_logtotrefh_c=s_logtotrefh-2.708

gen s_logtotrefhaidpercap=s_logtotrefh_c*(aidpercap_c*-1)

*model with interaction

logistic onset s_logtotrefh_c aidpercap_c s_logtotrefhaidpercap totpopln

polity politysq polmiss imr brevpeace if !(onset==0 & incidence==1) &

year==2001

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

94

*(2) CROSS-SECTIONAL INTERACTION S&G CONVENTIONAL MODEL 2.4. (potential

cropland & share of rural population)

*generate interaction

sum u_popdprodln

gen u_popdprodln_c=u_popdprodln-4.177

sum wup_prosrur

gen wup_prosrur_c=wup_prosrur-55.223

gen prosrurpopdprod=(u_popdprodln_c*-1)*wup_prosrur_c

*model with interaction

logistic onset s_logtotrefh prosrurpopdprod u_popdprodln_c wup_prosrur_c

imr totpopln polity politysq polmiss brevpeace if !(onset==0 &

incidence==1) & year==2001

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(3) CROSS-SECTIONAL MODEL CSP NON-STATE 3.1. (refugees & share of rural

population)

*generate interaction

sum csp_loghost

gen csp_loghost_c=csp_loghost-1.655

sum wup_prosrur

gen wup_prosrur_c=wup_prosrur-55.223

gen csp_loghostprosrur=csp_loghost_c*wup_prosrur_c

*model with interaction

logistic nonscinc csp_loghost_c csp_loghostprosrur wup_prosrur_c imr

totpopln polity politysq polmiss brevpeace if year==2007

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(3) CROSS-SECTIONAL MODEL CSP NON-STATE 3.2. (refugees & potential

cropland)

*generate interaction

sum csp_loghost

gen csp_loghost_c=csp_loghost-1.655

sum u_popdprodln

gen u_popdprodln_c=u_popdprodln-4.177

gen csp_loghostpopdprod=csp_loghost_c*u_popdprodln_c

*model with interaction

logistic nonscinc csp_loghost_c u_popdprodln_c csp_loghostpopdprod totpopln

polity politysq polmiss imr brevpeace if year==2007

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(3) CROSS-SECTIONAL MODEL CSP NON-STATE 3.3. (refugees & aid)

*generate interaction

95

sum aidpercap

gen aidpercap_c=aidpercap-27.697

sum csp_loghost

gen csp_loghost_c=csp_loghost-1.655

gen csp_loghostaidpercap=csp_loghost_c*(aidpercap_c*-1)

*model with interaction

logistic nonscinc csp_loghost_c aidpercap_c csp_loghostaidpercap totpopln

polity politysq polmiss imr brevpeace if year==2007

**************************************************************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(3) CROSS-SECTIONAL MODEL CSP NON-STATE 3.4. (potential cropland & share

of rural population)

*generate interaction

sum u_popdprodln

gen u_popdprodln_c=u_popdprodln-4.177

sum wup_prosrur

gen wup_prosrur_c=wup_prosrur-55.223

gen prosrurpopdprod=(u_popdprodln_c*-1)*wup_prosrur_c

*model with interaction

logistic nonscinc csp_loghost prosrurpopdprod u_popdprodln_c wup_prosrur_c

imr totpopln polity politysq polmiss brevpeace if year==2007

**************************************************************************

**************************************************************************

********************** DESCRIPTIVE STATISTICS ****************************

sum onset if !(onset==0 & incidence==1)

sum nonscinc

**************************************************************************

**************************************************************************

****************************** POST 1999 *********************************

clear

set mem 100m

use "M:\MINE DOKUMENTER\MASTER\final\odamainEd.dta"

*(1) LOGISTIC BASIC MODEL 1.0. (conventional conflict, CSP ref data)

logistic onset csp_loghost imr totpopln polity politysq polmiss brevpeace

if !(onset==0 & incidence==1) & year > 1999

***

*(1) LOGISTIC ALTERNATIVE MODEL

*Contecstual variables:

logistic onset csp_loghost u_popdprodln aidpercap wup_prosrur imr totpopln

polity politysq polmiss brevpeace if !(onset==0 & incidence==1) & year >

1999

*(2) LOGISTIC BASIC MODEL 2.0. (conventional conflict, Salehyan and

Gleditsch ref data)

96

logistic onset s_logtotrefh imr totpopln polity politysq polmiss brevpeace

if !(onset==0 & incidence==1) & year > 1999

***

*(2) LOGISTIC ALTERNATIVE MODEL

*Contecstual variables:

logistic onset s_logtotrefh u_popdprodln aidpercap wup_prosrur imr totpopln

polity politysq polmiss brevpeace if !(onset==0 & incidence==1) & year >

1999